Sam's Notebook https://robertslab.github.io/sams-notebook/ quarto-1.6.42 Tue, 08 Sep 2026 07:00:00 GMT Gape Monitoring - Initial Configuration and Simulations Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-09-08-Gape-Monitoring---Initial-Configuration-and-Simulations/

INTRO

A while ago, Steven asked me to spec out a gape monitoring system (GitHub Issue). The idea was based on the valve gape monitoring system (ScienceDirect) described in the paper by H. M. Stokes et al., 2024. In that paper, they set up a small Arduino-based system to monitor the valve gape of bivalves. The system used a small data logger, magnets and a three-axis magnetometer to measure assess frequency, duration, and the amplitude of the gape movements.

A primary advantage of this system is that it is relatively inexpensive (for example, the Arduino units only cost ~$13 each) and can be easily deployed in the field.

We’ve purchased something similar to the system described in the paper, and I was tasked with setting up the initial configuration and running some simulations to test the system’s performance.

MATERIALS & METHODS

I used the following materials to set up the initial configuration of the gape monitoring system:

  • Adafruit RP2040 Feather Adalogger (Arduino-based data logger)

  • Adafruit TLV493D 3-Axis Magnetometer

  • 4mm x 2mm neodymium magnets

  • one dead oyster

Initial Configuration

First, I installed the CircuitPython extension for VS Code. The primary advantage of this is being able to view/use the Serial console directly in VS Code, which makes it easier to debug and test the code.

The CircuitPython extension for VS Code is available in the VS Code marketplace. Briefly, once it’s installed, connect the Adafruit RP2040 Feather Adalogger to your computer via USB, and then Open the the CICRUITPY folder in VS Code. The CircuitPython extension will automatically detect the Adafruit RP2040 Feather Adalogger. Then, you can open the Serial console in VS Code by clicking on the “CircuitPython: Open Serial Console” button in the bottom right corner of VS Code. This will open a terminal window that allows you to view the output from the Adafruit RP2040 Feather Adalogger.

The Adfruit website has a great tutorial (Adafruit website) on how to set up the Adafruit RP2040 Feather Adalogger with CircuitPython. I followed the steps in that tutorial to get the data logger up and running.

The initial configuration involved the following steps:

  1. Initialize the Adafruit RP2040 Feather Adalogger with CircuitPython (Adafruit website).

  2. Install the CircuitPython libraries, using the entire Adafruit CircuitPython Bundle for Version 10.x, on the Adafruit RP2040 Feather Adalogger.

  3. Attach the Adafruit TLV493D 3-Axis Magnetometer to the Adafruit RP2040 Feather Adalogger via the STEMMA QT connection.

After that was setup, I used some basic code provided by Adafruit to test the magnetometer and ensure that I could get readings. The code was able to read the magnetic field values from the magnetometer and print them to the console.

Code

With the assistance Claude Science, I was able to modify the code to log the magnetic field values to a CSV file on the Adafruit RP2040 Feather Adalogger. The code is available in code.py (GitHub). The code logs the magnetic field values to a CSV file every 1 second. The CSV file is saved to the Adafruit RP2040 Feather Adalogger’s SD card. Currently, the code only logs cumulative time elapsed since the start of the program (e.g. when the Adafruit RP2040 Feather Adalogger is powered on) due to the fact that the Adafruit RP2040 Feather Adalogger does not have a real-time clock. However, I plan to add a real-time clock to the system in the future, which will allow me to log the actual date and time of each reading.

Important

The code will be continually updated and improved over time. Please check the GitHub repository for the latest version of the code. For example, currently the code does not perform any calibration adjustments for the magnetometer and does not assess gape distance. Additionally, the code does not assess whether or not the changes in the magnetic field values identified as gape events are statistically significantly different from the background noise. These are all things that will be added in the future.

Simulations

I used masking tape to attache the magnetometer to the leading edge of the bottom valve of the dead oyster. Initially, I attached a single magnet to the leading edge of the top valve of the dead oyster. The magnetic field that was able to be detected from a single magnet was very weak and had me concerned that they would be indistinguishable from the background noise. I then stacked (side-by-side seemed to cancel out the magnetic field) a second magnet to the leading edge of the top valve of the dead oyster, and this increased the magnetic field strength that was detected by the magnetometer.

I used small piece of plastic to keep the shell open ~5mm for various durations and recorded the data from the magnetometer. The data was logged to a CSV file on the Adafruit RP2040 Feather Adalogger’s SD card. I then used R to read in the CSV file and plot the magnetic field values over time.

Pacific oyster with green masking tape on the leading edge of the bottom valve to hold magnetometer, and two neodymium magnets stacked on the leading edge of the top valve. The light on the RP2040 Adalogger is illuminated, indicating it’s capturing data at that moment.

Analysis

Using Claude Science, I produced an R Markdown file (GitHub) was able to analyze the data from the CSV file and plot the magnetic field values over time, while also highlighting the periods of time when the shell was open.

RESULTS

Data analysis seems to work as desired. The magnetic field values are clearly distinguishable from the background noise, and the periods of time when the shell was open are clearly visible in the plots.

Magnetic field magnitude over time shown for three data sets. The 3rd data set spans ~6000 seconds and highlights the periods when the shell was open in grey.

SUMMARY

Overall, this is a promising start to the gape monitoring system. The initial configuration and simulations have shown that the system is capable of detecting the valve gape of bivalves, and the data analysis has been successful in identifying the periods of time when the shell was open.

Next, I plan to add a real-time clock to the system, which will allow me to log the actual date and time of each reading. Additionally, I plan to determine calibration values (e.g. correlate magnetic field values with known gape measurements) for all three magnetometers, which will allow me to convert the magnetic field values to actual gape measurements. Once these are completed, I’ll test out on live oysters in the lab.

]]> 2026 https://robertslab.github.io/sams-notebook/posts/2026/2026-09-08-Gape-Monitoring---Initial-Configuration-and-Simulations/ Tue, 08 Sep 2026 07:00:00 GMT Resazurin Assays - USDA M.gigas Juveniles OAE Preconditioning Effect on 36C Heat Stress Response Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-31-Resazurin-Assays---USDA-M.gigas-Juveniles-OAE-Preconditioning-Effect-on-36C-Heat-Stress-Response/

INTRO

Steven wanted to assess metabolism of juvenile Magallana gigas (Pacific oyster) after ocean alkalinity enhancement (OAE) preconditioning and subsequent heat stress. The goal was to determine whether OAE preconditioning would affect metabolic response to 36°C heat stress.

MATERIALS & METHODS

Juvenile Magallana gigas (Pacific oyster) were subjected to OAE preconditioning at 10°C for 6 hours on 20260828. After exposure, oysters were returned to standard seawater conditions. This was performed by Steven.

Today, 46 OAE-preconditioned oysters and 46 control oysters were subjected to 36°C heat stress. Oysters were randomly distributed across four 48-well clear plates and submerged in 2mL of resazurin working solution, with fluorescence measured every 30 minutes in a Synergy HTX plate reader (Agilent).

Resazurin working solution was prepared from a stock solution made by me on 20260610. The working solution was made with Instant Ocean held at 10°C.

Reagent Volume
Instant Ocean 208mL
resazurin stock solution 469uL
DMSO 211uL
100x Penn/Strep & 100x Fungizone 2.11mL

Resazurin working-solution temperatures were recorded by the plate reader at each read (mean across the four plates):

Timepoint (hrs) Temp (C)
0 22.2
0.5 23.0
1 23.7
1.5 24.4
2 24.7
2.5 25.1
3 25.5

RESULTS

Treatment significantly affected size-normalized metabolic rate (AUC ANOVA: F = 4.33, p = 0.0403). OAE-preconditioned oysters had higher mean metabolic activity than controls (Tukey: estimate = -9e-04, p = 0.0403).

Markdown below was knitted from 01.00-resazurin-20260831-mgig-OAE-36C.Rmd (GitHub).

AI Use Level 4: Substantial AI contribution

1 Background

Juvenile oysters were subjected to OAE conditions (10°C) for 6hrs on 20260828. After exposure, oysters were transferred to standard sea water conditions. On 20260831, 46 OAE oysters and 46 control oysters were subjected to 36°C heat stress. Oysters were randomly distributed across four, 48-well clear plates and submerged in 2mL of resazurin working solution. Fluorescence was measured every 30mins in a Synergy HTX (Agilent) plate reader.

Resazurin working-solution temperatures were recorded by the plate reader at each read (mean across the four plates):

Timepoint (hrs) Temp (C)
0 22.2
0.5 23.0
1 23.7
1.5 24.4
2 24.7
2.5 25.1
3 25.5

See Resazurin/data/20260831-mgig-OAE-36C/README.md for full experimental notes.

1.1 Expected inputs

Path Description
Resazurin/data/20260831-mgig-OAE-36C/plate-*-T*.txt Plate reader fluorescence exports (one file per plate per timepoint)
Resazurin/data/20260831-mgig-OAE-36C/layout.csv Well metadata: plate ID, well ID, blank flag, treatment group (OAE vs. control), sample IDs, area measurements (mm², from ImageJ)

1.2 Expected outputs

All outputs are written to Resazurin/outputs/01.00-resazurin-20260831-mgig-OAE-36C/.

File Description
figures/ All plots generated by this script
auc_all_metrics.csv Per-individual AUC values for every active measurement metric
auc_summary.csv Group-level AUC summary statistics (mean, SD, SE, median)
metabolism.csv Full per-well per-timepoint metabolism data frame
pairwise_stats.csv Tukey-adjusted pairwise comparisons from AUC linear models

2 Setup

2.1 Knitr options

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,        # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",         # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Load libraries

library(tidyverse)
library(pracma)       # trapz()
library(lme4)
library(lmerTest)
library(emmeans)
library(multcompView)
library(cowplot)
library(colorspace)   # qualitative_hcl() for large palettes

3 Helper Functions

normalize_well_id <- function(x) {
  x <- toupper(trimws(x))
  valid <- str_detect(x, "^[A-Z]+[0-9]+$")
  out <- rep(NA_character_, length(x))
  if (!any(valid)) return(out)
  m <- str_match(x[valid], "^([A-Z]+)([0-9]+)$")
  out[valid] <- paste0(m[, 2], as.integer(m[, 3]))
  out
}

parse_time_hr <- function(path) {
  hit <- str_match(basename(path),
                   "(?i)-T([0-9]+(?:\\.[0-9]+)?)\\.txt$")
  as.numeric(hit[, 2])
}

parse_plate_id <- function(path) {
  hit <- str_match(basename(path),
    "(?i)^plate-([A-Za-z0-9-]+)-T[0-9]+(?:\\.[0-9]+)?\\.txt$")
  id <- hit[, 2]
  ifelse(is.na(id), "unknown", id)
}

extract_results_block <- function(lines) {
  results_idx <- which(trimws(lines) == "Results")
  if (length(results_idx) == 0) stop("No Results section found")
  idx <- results_idx[1]
  header_tokens <- str_split(lines[idx + 1], "\\t")[[1]] |> trimws()
  col_ids <- header_tokens[
    header_tokens != "" & str_detect(header_tokens, "^[0-9]+$")]
  j <- idx + 2
  data_lines <- character()
  while (j <= length(lines)) {
    line <- lines[j]
    if (trimws(line) == "") break
    if (!str_detect(line, "^[A-Za-z]\\t")) break
    data_lines <- c(data_lines, line)
    j <- j + 1
  }
  list(col_ids = col_ids, data_lines = data_lines)
}

parse_plate_export <- function(path) {
  lines <- readLines(path, warn = FALSE)
  res <- extract_results_block(lines)

  map_dfr(res$data_lines, function(line) {
    tokens <- str_split(line, "\\t")[[1]] |> trimws()
    tokens <- tokens[tokens != ""]
    row_letter <- tokens[1]
    nums <- suppressWarnings(as.numeric(tokens[-1]))
    valid_idx <- which(!is.na(nums))
    if (length(valid_idx) == 0) return(tibble())
    vals <- nums[valid_idx]
    n <- min(length(vals), length(res$col_ids))
    tibble(
      row_id  = toupper(row_letter),
      col_id  = as.integer(res$col_ids[seq_len(n)]),
      well_id = normalize_well_id(
        paste0(toupper(row_letter), res$col_ids[seq_len(n)])),
      value   = vals[seq_len(n)]
    )
  }) %>%
    mutate(
      plate_id = str_to_lower(parse_plate_id(path)),
      time_hr  = parse_time_hr(path)
    )
}

trapezoid_auc <- function(time_hr, value) {
  ok <- is.finite(time_hr) & is.finite(value)
  t <- time_hr[ok]
  v <- value[ok]
  if (length(t) < 2) return(NA_real_)
  ord <- order(t)
  t <- t[ord]; v <- v[ord]
  sum(diff(t) * (head(v, -1) + tail(v, -1)) / 2)
}

# Shared helper: extract display unit string from a measurement column name.
# e.g. "area_mm2_measurement" -> "mm²", "weight_mg_measurement" -> "mg"
parse_meas_unit <- function(col_name) {
  unit_raw <- col_name |>
    str_remove("^metabolism_per_") |>
    str_remove("_measurement$") |>
    str_extract("[^_]+$")
  case_when(
    unit_raw == "mm2" ~ "mm²",
    unit_raw == "cm2" ~ "cm²",
    unit_raw == "mm3" ~ "mm³",
    unit_raw == "cm3" ~ "cm³",
    TRUE              ~ unit_raw
  )
}

# y-axis label for metabolism line plots: "fold change/mm²"
metabolism_y_label <- function(col_name) {
  paste0("Metabolism (fold change/", parse_meas_unit(col_name), ")")
}

# y-axis label for AUC box plots: "Metabolism (AUC; mm²)"
auc_y_label <- function(metric_name) {
  paste0("Metabolism (AUC; ", parse_meas_unit(metric_name), ")")
}

4 Load Data

4.1 Plate export files

proj_root <- rprojroot::find_rstudio_root_file()
data_dir  <- file.path(proj_root, "Resazurin", "data", "20260831-mgig-OAE-36C")
out_dir   <- file.path(proj_root, "Resazurin", "outputs",
                        "01.00-resazurin-20260831-mgig-OAE-36C")
fig_dir   <- file.path(out_dir, "figures")

dir.create(fig_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)

plate_files <- list.files(
  data_dir,
  pattern = "(?i)^plate-.*-T[0-9]+(?:\\.[0-9]+)?\\.txt$",
  full.names = TRUE
)

plate_raw <- map_dfr(plate_files, function(path) {
  tryCatch(parse_plate_export(path),
           error = function(e) {
             message("Parse error in ", basename(path), ": ", e$message)
             tibble()
           })
})

str(plate_raw)
tibble [672 × 6] (S3: tbl_df/tbl/data.frame)
 $ row_id  : chr [1:672] "A" "A" "A" "A" ...
 $ col_id  : int [1:672] 1 2 3 4 5 6 1 2 3 4 ...
 $ well_id : chr [1:672] "A1" "A2" "A3" "A4" ...
 $ value   : num [1:672] 202 195 244 211 178 222 204 187 232 207 ...
 $ plate_id: chr [1:672] "a" "a" "a" "a" ...
 $ time_hr : num [1:672] 0 0 0 0 0 0 0 0 0 0 ...

4.2 Plate consistency check

Checks that every plate has the same number of wells at every timepoint. The expected well count is the mode across all plate × timepoint reads. Any plate with at least one deviating read is flagged and dropped entirely before any further analysis — removing only the aberrant timepoint would break the fold-change baseline calculation.

well_counts <- plate_raw %>%
  group_by(plate_id, time_hr) %>%
  summarise(n_wells = n_distinct(well_id), .groups = "drop")

expected_n_wells <- as.integer(
  names(which.max(table(well_counts$n_wells)))
)

inconsistent_reads <- well_counts %>%
  filter(n_wells != expected_n_wells) %>%
  arrange(plate_id, time_hr)

inconsistent_plate_ids <- unique(inconsistent_reads$plate_id)

if (nrow(inconsistent_reads) > 0) {
  cat("**Plate consistency check FAILED.**",
      "Expected", expected_n_wells, "wells per plate-timepoint read.",
      length(inconsistent_plate_ids),
      "plate(s) have at least one deviating read and are excluded",
      "from all analyses:\n\n")
  cat(knitr::kable(
    inconsistent_reads,
    col.names = c("Plate", "Time (h)", "Wells read"),
    caption   = paste("Expected:", expected_n_wells, "wells per read")
  ), sep = "\n")
  cat("\n")
  plate_raw <- plate_raw %>%
    filter(!plate_id %in% inconsistent_plate_ids)
  message(length(inconsistent_plate_ids),
          " plate(s) removed from plate_raw: ",
          paste(inconsistent_plate_ids, collapse = ", "))
} else {
  cat("Plate consistency check passed: all",
      n_distinct(well_counts$plate_id), "plates have",
      expected_n_wells, "wells at every timepoint.\n")
}

Plate consistency check passed: all 4 plates have 24 wells at every timepoint.

4.3 Layout file

layout_path <- file.path(data_dir, "layout.csv")

layout_raw <- read_csv(layout_path,
                       col_types = cols(.default = "c"),
                       show_col_types = FALSE)

# Standardise column names to snake_case
names(layout_raw) <- names(layout_raw) |>
  str_to_lower() |>
  str_replace_all("[^a-z0-9]+", "_") |>
  str_replace_all("_+", "_") |>
  str_replace("_$", "")

# Normalise plate_id to match plate file ids (strip "plate-" prefix)
layout_clean <- layout_raw %>%
  mutate(
    plate_id = str_remove(str_to_lower(plate_id), "^plate-"),
    well_id  = normalize_well_id(plate_well),
    is_blank = if ("is_blank" %in% names(layout_raw))
      toupper(trimws(is_blank)) %in% c("TRUE", "T", "1", "YES", "Y")
    else
      FALSE
  )

found_exclude_col <- intersect(
  c("exclude_from_analysis", "exclude", "omit", "not_analyzed"),
  names(layout_clean)
)[1]
layout_clean <- layout_clean %>%
  mutate(
    exclude_from_analysis = if (!is.na(found_exclude_col))
      toupper(trimws(.data[[found_exclude_col]])) %in%
        c("TRUE", "T", "1", "YES", "Y")
    else
      FALSE
  )

# Identify measurement columns and group columns
measurement_cols <- names(layout_clean)[
  str_detect(names(layout_clean), "_measurement$")]
group_cols <- names(layout_clean)[
  str_detect(names(layout_clean), "_group$")]

# Cast measurement columns to numeric
layout_clean <- layout_clean %>%
  mutate(across(all_of(measurement_cols),
                ~ suppressWarnings(as.numeric(.x))))

# Determine which measurement columns actually contain finite data
active_meas_cols <- measurement_cols[
  sapply(measurement_cols, function(col)
    any(is.finite(layout_clean[[col]]), na.rm = TRUE))]

# Normalise group values to lowercase so they match colour scale definitions
layout_clean <- layout_clean %>%
  mutate(across(all_of(group_cols),
                ~ str_to_lower(trimws(as.character(.x)))))

message("Group columns: ", paste(group_cols, collapse = ", "))
message("Active measurement columns: ",
        paste(active_meas_cols, collapse = ", "))

str(layout_clean)
tibble [96 × 14] (S3: tbl_df/tbl/data.frame)
 $ plate_id             : chr [1:96] "a" "a" "a" "a" ...
 $ plate_well           : chr [1:96] "A01" "A02" "A03" "A04" ...
 $ is_blank             : logi [1:96] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ family_id_group      : chr [1:96] NA NA NA NA ...
 $ sample_id_group      : chr [1:96] "c-01" "c-25" "oae-03" "oae-27" ...
 $ treatment_group      : chr [1:96] "control" "control" "oae" "oae" ...
 $ exclude_from_analysis: logi [1:96] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:96] NA NA NA NA ...
 $ width_mm_measurement : num [1:96] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:96] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:96] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:96] 7395 7340 6008 8749 8809 ...
 $ imagej_id            : chr [1:96] "8" "24" "1" "39" ...
 $ well_id              : chr [1:96] "A1" "A2" "A3" "A4" ...

5 Merge Plate Data with Layout

dat <- plate_raw %>%
  left_join(
    layout_clean %>%
      select(plate_id, well_id, is_blank, exclude_from_analysis,
             any_of("exclude_reason"),
             all_of(group_cols), all_of(measurement_cols)),
    by = c("plate_id", "well_id")
  ) %>%
  mutate(
    is_blank = replace_na(is_blank, FALSE),
    exclude_from_analysis = replace_na(exclude_from_analysis, FALSE)
  )

str(dat)
tibble [672 × 16] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:672] "A" "A" "A" "A" ...
 $ col_id               : int [1:672] 1 2 3 4 5 6 1 2 3 4 ...
 $ well_id              : chr [1:672] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:672] 202 195 244 211 178 222 204 187 232 207 ...
 $ plate_id             : chr [1:672] "a" "a" "a" "a" ...
 $ time_hr              : num [1:672] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:672] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:672] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:672] NA NA NA NA ...
 $ family_id_group      : chr [1:672] NA NA NA NA ...
 $ sample_id_group      : chr [1:672] "c-01" "c-25" "oae-03" "oae-27" ...
 $ treatment_group      : chr [1:672] "control" "control" "oae" "oae" ...
 $ width_mm_measurement : num [1:672] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:672] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:672] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:672] 7395 7340 6008 8749 8809 ...

6 Raw Fluorescence

6.1 Data frame

# Wells in the plate reader output that have no layout entry get all-NA group
# columns after the join. Keep only wells assigned to at least one group.
active_gc <- intersect(group_cols, names(dat))

raw_df <- dat %>%
  filter(
    !is_blank,
    if (length(active_gc) > 0)
      if_any(all_of(active_gc), ~ !is.na(.))
    else
      TRUE
  ) %>%
  mutate(
    trace_id = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  )

families   <- str_sort(unique(na.omit(raw_df$family_id_group)), numeric = TRUE)
treatments <- sort(unique(na.omit(raw_df$treatment_group)))

n_fam <- length(families)
n_trt <- length(treatments)

# Palette strategy:
#   <= 7 groups : Okabe-Ito (gold standard for colorblind-safe figures).
#   >  7 groups : colorspace::qualitative_hcl("Dynamic") scales to any N
#                 using perceptually uniform HCL space — no colour collisions.
# Black (#000000) is excluded from both and reserved for blank wells.
okabe_ito_7 <- c(
  "#E69F00", "#56B4E9", "#009E73", "#F0E442",
  "#0072B2", "#D55E00", "#CC79A7"
)
make_palette <- function(n) {
  if (n == 0L) return(character(0))
  if (n <= length(okabe_ito_7)) return(okabe_ito_7[seq_len(n)])
  colorspace::qualitative_hcl(n, palette = "Dynamic")
}

all_colours   <- make_palette(n_fam + n_trt)
fam_colours   <- setNames(all_colours[seq_len(n_fam)], families)
trt_colours   <- setNames(all_colours[n_fam + seq_len(n_trt)], treatments)

lty_pool <- c("solid", "dashed", "dotted", "dotdash", "longdash")
trt_linetypes <- setNames(
  lty_pool[(seq_len(n_trt) - 1L) %% length(lty_pool) + 1L],
  treatments
)
plate_well_colours <- c(blank = "black", fam_colours)

has_trt <- n_trt > 0
has_fam <- n_fam > 0

str(raw_df)
tibble [644 × 17] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:644] "A" "A" "A" "A" ...
 $ col_id               : int [1:644] 1 2 3 4 5 6 1 2 3 4 ...
 $ well_id              : chr [1:644] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:644] 202 195 244 211 178 222 204 187 232 207 ...
 $ plate_id             : chr [1:644] "a" "a" "a" "a" ...
 $ time_hr              : num [1:644] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:644] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:644] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:644] NA NA NA NA ...
 $ family_id_group      : chr [1:644] NA NA NA NA ...
 $ sample_id_group      : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ treatment_group      : chr [1:644] "control" "control" "oae" "oae" ...
 $ width_mm_measurement : num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:644] 7395 7340 6008 8749 8809 ...
 $ trace_id             : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...

6.2 Raw fluorescence by plate (including blanks)

# Colour wells by family when family groups exist; otherwise fall back to
# treatment, or a flat "sample" bucket when neither grouping is populated.
plate_group_col <- if (has_fam) "family_id_group" else if (has_trt) "treatment_group" else NULL
plate_group_colours <- if (has_fam) c(blank = "black", fam_colours) else if (has_trt) c(blank = "black", trt_colours) else c(blank = "black", sample = "grey40")
plate_group_name <- if (has_fam) "Family" else if (has_trt) "Treatment" else "Group"

p_raw_plates <- dat %>%
  filter(is.finite(time_hr), is.finite(value)) %>%
  mutate(
    colour_group = if_else(
      is_blank, "blank",
      if (!is.null(plate_group_col)) coalesce(.data[[plate_group_col]], "sample")
      else "sample"
    ),
    trace_id     = paste(plate_id, well_id, sep = "_")
  ) %>%
  ggplot(aes(x = time_hr, y = value,
             group = trace_id, colour = colour_group)) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1, alpha = 0.7) +
  facet_wrap(~ plate_id) +
  scale_colour_manual(
    values   = plate_group_colours,
    name     = plate_group_name,
    breaks   = names(plate_group_colours),
    na.value = "grey80"
  ) +
  labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_raw_plates

ggsave(file.path(fig_dir, "raw_fluor_by_plate.png"),
       p_raw_plates, width = 10, height = 8)
if (has_fam) cat("## Mean raw fluorescence by family\n\n")
if (has_fam) {
  raw_family_summary <- raw_df %>%
    filter(!is.na(family_id_group), !exclude_from_analysis) %>%
    group_by(family_id_group, treatment_group, time_hr) %>%
    summarise(
      mean_fluor = mean(value, na.rm = TRUE),
      se_fluor   = sd(value, na.rm = TRUE) /
        sqrt(sum(!is.na(value))),
      n          = sum(!is.na(value)),
      .groups    = "drop"
    ) %>%
    mutate(group_var = if (has_trt)
      paste(family_id_group, treatment_group, sep = ".")
    else
      family_id_group)

  p_raw_mean <- ggplot(raw_family_summary,
      aes(x = time_hr, y = mean_fluor,
          colour = family_id_group,
          group = group_var)) +
    geom_ribbon(aes(ymin = mean_fluor - se_fluor,
                    ymax = mean_fluor + se_fluor,
                    fill = family_id_group),
                alpha = 0.15, colour = NA) +
    geom_line(
      mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
      linewidth = 1) +
    geom_point(size = 2) +
    scale_colour_manual(values = fam_colours, name = "Family",
                        breaks = families) +
    scale_fill_manual(values = fam_colours, name = "Family",
                      breaks = families) +
    labs(x = "Time (h)", y = "Mean raw fluorescence (RFU ± SE)") +
    theme_classic(base_size = 13) +
    if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

  p_raw_mean
  ggsave(file.path(fig_dir, "raw_mean_by_family.png"),
         p_raw_mean, width = 8, height = 5)
} else {
  message("No family groups present: skipping mean raw fluorescence by family.")
}
if (has_fam) cat("## Individual raw fluorescence traces by family\n\n")
if (has_fam) {
  p_raw_by_family <- raw_df %>%
    filter(!is.na(family_id_group)) %>%
    ggplot(aes(x = time_hr, y = value, group = trace_id,
               colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ family_id_group) +
    scale_colour_manual(
      values = if (has_trt) trt_colours else fam_colours,
      name   = if (has_trt) "Treatment" else "Family",
      breaks = if (has_trt) treatments else families) +
    labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_raw_by_family
  ggsave(file.path(fig_dir, "raw_individual_by_family.png"),
         p_raw_by_family, width = 10, height = 5)
} else {
  message("No family groups present: skipping individual raw fluorescence traces by family.")
}

6.3 Individual raw fluorescence traces by treatment

if (has_trt) {
  colour_var   <- if (has_fam) "family_id_group" else "treatment_group"
  colour_vals  <- if (has_fam) fam_colours else trt_colours
  colour_brks  <- if (has_fam) families else treatments
  colour_name  <- if (has_fam) "Family" else "Treatment"

  p_raw_by_treatment <- raw_df %>%
    ggplot(aes(x = time_hr, y = value,
               group = trace_id, colour = .data[[colour_var]])) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ treatment_group) +
    scale_colour_manual(values = colour_vals, name = colour_name,
                        breaks = colour_brks) +
    labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_raw_by_treatment
  ggsave(file.path(fig_dir, "raw_individual_by_treatment.png"),
         p_raw_by_treatment, width = 10, height = 5)
}

6.4 Excluded samples

Wells flagged exclude_from_analysis = TRUE appear in the raw fluorescence plots above but are omitted from all analyses that follow.

excl_cols <- c("plate_id", "well_id", "sample",
               if (has_fam) "family_id_group",
               if (has_trt) "treatment_group")

excluded_wells <- dat %>%
  filter(!is_blank, exclude_from_analysis) %>%
  mutate(
    sample = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  ) %>%
  select(all_of(excl_cols), any_of("exclude_reason")) %>%
  distinct() %>%
  arrange(plate_id, well_id)

if (nrow(excluded_wells) > 0) {
  col_names <- c("Plate", "Well", "Sample",
                 if (has_fam) "Family",
                 if (has_trt) "Treatment")
  if ("exclude_reason" %in% names(excluded_wells))
    col_names <- c(col_names, "Reason")
  cat(knitr::kable(excluded_wells, col.names = col_names), sep = "\n")
} else {
  cat("No wells are excluded from analysis.\n")
}

No wells are excluded from analysis.

7 Blank Correction via Fold-Change Normalization

T0 is the earliest timepoint present in the dataset (not necessarily 0 hr). Sample fold-change is expressed relative to each individual’s T0 reading, resolved by sample_id_group when that column is populated — allowing the same animal to be tracked across plates — or by plate_id + well_id when no sample IDs exist (backward-compatible with single-plate, multi-timepoint designs). Blank fold-change is the per-plate mean blank RFU at each timepoint divided by the pooled mean blank RFU at T0. Subtracting blank fold-change from sample fold-change removes background fluorescence drift; all samples start at exactly 0 at T0 by construction.

7.1 Step 1 – Identify T0 and compute per-sample fold-change

# T0 = earliest timepoint present in the dataset
t0_time <- min(dat$time_hr[is.finite(dat$time_hr)], na.rm = TRUE)
message("T0 timepoint: ", t0_time, " hr")

# T0 reference value per individual.
# Resolved by sample_id_group (cross-plate tracking) when available;
# falls back to plate+well for layouts without explicit sample IDs.
t0_all <- dat %>%
  filter(time_hr == t0_time, !is_blank, is.finite(value)) %>%
  mutate(sample_key = if_else(
    !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
    as.character(sample_id_group),
    paste(plate_id, well_id, sep = "_")
  )) %>%
  group_by(sample_key) %>%
  summarise(value_t0 = mean(value, na.rm = TRUE), .groups = "drop")

dat_fc <- dat %>%
  mutate(sample_key = if_else(
    !is_blank &
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
    as.character(sample_id_group),
    paste(plate_id, well_id, sep = "_")
  )) %>%
  left_join(t0_all, by = "sample_key") %>%
  mutate(fold_change = if_else(
    !is_blank & is.finite(value_t0) & value_t0 > 0,
    value / value_t0,
    NA_real_
  ))

str(dat_fc)
tibble [672 × 19] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:672] "A" "A" "A" "A" ...
 $ col_id               : int [1:672] 1 2 3 4 5 6 1 2 3 4 ...
 $ well_id              : chr [1:672] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:672] 202 195 244 211 178 222 204 187 232 207 ...
 $ plate_id             : chr [1:672] "a" "a" "a" "a" ...
 $ time_hr              : num [1:672] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:672] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:672] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:672] NA NA NA NA ...
 $ family_id_group      : chr [1:672] NA NA NA NA ...
 $ sample_id_group      : chr [1:672] "c-01" "c-25" "oae-03" "oae-27" ...
 $ treatment_group      : chr [1:672] "control" "control" "oae" "oae" ...
 $ width_mm_measurement : num [1:672] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:672] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:672] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:672] 7395 7340 6008 8749 8809 ...
 $ sample_key           : chr [1:672] "c-01" "c-25" "oae-03" "oae-27" ...
 $ value_t0             : num [1:672] 202 195 244 211 178 222 204 187 232 207 ...
 $ fold_change          : num [1:672] 1 1 1 1 1 1 1 1 1 1 ...

7.2 Step 2 – Blank fold-change reference per plate per timepoint

# Pooled mean blank RFU at T0 across all T0 plates
mean_blank_t0 <- dat %>%
  filter(is_blank, time_hr == t0_time, is.finite(value)) %>%
  pull(value) %>%
  mean(na.rm = TRUE)

if (!is.finite(mean_blank_t0))
  message("No blank readings found at T0 (", t0_time,
          " hr); blank correction will produce NA.")

# Per-plate per-timepoint mean blank expressed as fold-change relative to T0
blank_fc_ref <- dat %>%
  filter(is_blank, is.finite(value)) %>%
  group_by(plate_id, time_hr) %>%
  summarise(mean_blank_rfu = mean(value, na.rm = TRUE), .groups = "drop") %>%
  mutate(mean_blank_fc = mean_blank_rfu / mean_blank_t0)

str(blank_fc_ref)
tibble [28 × 4] (S3: tbl_df/tbl/data.frame)
 $ plate_id      : chr [1:28] "a" "a" "a" "a" ...
 $ time_hr       : num [1:28] 0 0.5 1 1.5 2 2.5 3 0 0.5 1 ...
 $ mean_blank_rfu: num [1:28] 168 167 169 171 172 172 173 161 165 165 ...
 $ mean_blank_fc : num [1:28] 0.973 0.967 0.978 0.99 0.996 ...

7.3 Step 3 – Subtract blank fold-change from sample fold-change

samples <- dat_fc %>%
  filter(!is_blank, !exclude_from_analysis) %>%
  mutate(
    trace_id = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  ) %>%
  left_join(blank_fc_ref, by = c("plate_id", "time_hr")) %>%
  mutate(corrected_fc = fold_change - mean_blank_fc)

str(samples)
tibble [644 × 23] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:644] "A" "A" "A" "A" ...
 $ col_id               : int [1:644] 1 2 3 4 5 6 1 2 3 4 ...
 $ well_id              : chr [1:644] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:644] 202 195 244 211 178 222 204 187 232 207 ...
 $ plate_id             : chr [1:644] "a" "a" "a" "a" ...
 $ time_hr              : num [1:644] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:644] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:644] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:644] NA NA NA NA ...
 $ family_id_group      : chr [1:644] NA NA NA NA ...
 $ sample_id_group      : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ treatment_group      : chr [1:644] "control" "control" "oae" "oae" ...
 $ width_mm_measurement : num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:644] 7395 7340 6008 8749 8809 ...
 $ sample_key           : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ value_t0             : num [1:644] 202 195 244 211 178 222 204 187 232 207 ...
 $ fold_change          : num [1:644] 1 1 1 1 1 1 1 1 1 1 ...
 $ trace_id             : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ mean_blank_rfu       : num [1:644] 168 168 168 168 168 168 168 168 168 168 ...
 $ mean_blank_fc        : num [1:644] 0.973 0.973 0.973 0.973 0.973 ...
 $ corrected_fc         : num [1:644] 0.0275 0.0275 0.0275 0.0275 0.0275 ...

8 Blank-Corrected Fold-Change

if (has_fam) cat("## Mean by family\n\n")
if (has_fam) {
  bc_fc_summary <- samples %>%
    filter(!is.na(family_id_group), !exclude_from_analysis) %>%
    group_by(family_id_group, treatment_group, time_hr) %>%
    summarise(
      mean_val = mean(corrected_fc, na.rm = TRUE),
      se_val   = sd(corrected_fc, na.rm = TRUE) /
        sqrt(sum(!is.na(corrected_fc))),
      n        = sum(!is.na(corrected_fc)),
      .groups  = "drop"
    ) %>%
    mutate(group_var = if (has_trt)
      paste(family_id_group, treatment_group, sep = ".")
    else
      family_id_group)

  p_bc_fc_mean <- ggplot(bc_fc_summary,
      aes(x = time_hr, y = mean_val,
          colour = family_id_group,
          group = group_var)) +
    geom_ribbon(aes(ymin = mean_val - se_val,
                    ymax = mean_val + se_val,
                    fill = family_id_group),
                alpha = 0.15, colour = NA) +
    geom_line(
      mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
      linewidth = 1) +
    geom_point(size = 2) +
    scale_colour_manual(values = fam_colours, name = "Family",
                        breaks = families) +
    scale_fill_manual(values = fam_colours, name = "Family",
                      breaks = families) +
    labs(x = "Time (h)",
         y = "Mean blank-corrected fold-change (± SE)") +
    theme_classic(base_size = 13) +
    if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

  p_bc_fc_mean
  ggsave(file.path(fig_dir, "blank_corrected_fc_mean_by_family.png"),
         p_bc_fc_mean, width = 8, height = 5)
} else {
  message("No family groups present: skipping mean blank-corrected fold-change by family.")
}
if (has_fam) cat("## Individual traces by family\n\n")
if (has_fam) {
  p_bc_fc_by_family <- samples %>%
    filter(!is.na(family_id_group)) %>%
    ggplot(aes(x = time_hr, y = corrected_fc, group = trace_id,
               colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ family_id_group) +
    scale_colour_manual(
      values = if (has_trt) trt_colours else fam_colours,
      name   = if (has_trt) "Treatment" else "Family",
      breaks = if (has_trt) treatments else families) +
    labs(x = "Time (h)", y = "Blank-corrected fold-change") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_bc_fc_by_family
  ggsave(file.path(fig_dir, "blank_corrected_fc_by_family.png"),
         p_bc_fc_by_family, width = 10, height = 5)
} else {
  message("No family groups present: skipping individual blank-corrected fold-change traces by family.")
}

8.1 Individual blank-corrected fold-change traces by treatment

if (has_trt) {
  colour_var  <- if (has_fam) "family_id_group" else "treatment_group"
  colour_vals <- if (has_fam) fam_colours else trt_colours
  colour_brks <- if (has_fam) families else treatments
  colour_name <- if (has_fam) "Family" else "Treatment"

  p_bc_fc_by_treatment <- samples %>%
    ggplot(aes(x = time_hr, y = corrected_fc,
               group = trace_id, colour = .data[[colour_var]])) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ treatment_group) +
    scale_colour_manual(values = colour_vals, name = colour_name,
                        breaks = colour_brks) +
    labs(x = "Time (h)", y = "Blank-corrected fold-change") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_bc_fc_by_treatment
  ggsave(file.path(fig_dir, "blank_corrected_fc_by_treatment.png"),
         p_bc_fc_by_treatment, width = 10, height = 5)
}

9 Metabolism (Size-Normalised Fold-Change)

Blank-corrected fold-change divided by each active measurement column. This is “metabolism” as defined in Huffmyer et al.

if (length(active_meas_cols) == 0) {
  message("No active measurement columns: skipping metabolism calculation.")
  metabolism_df <- tibble()
} else {
  metabolism_df <- samples
  for (mc in active_meas_cols) {
    out_col <- paste0("metabolism_per_", mc)
    metabolism_df <- metabolism_df %>%
      mutate(!!out_col := if_else(
        is.finite(.data[[mc]]) & .data[[mc]] > 0 &
          is.finite(corrected_fc),
        corrected_fc / .data[[mc]],
        NA_real_
      ))
  }
}

str(metabolism_df)
tibble [644 × 24] (S3: tbl_df/tbl/data.frame)
 $ row_id                             : chr [1:644] "A" "A" "A" "A" ...
 $ col_id                             : int [1:644] 1 2 3 4 5 6 1 2 3 4 ...
 $ well_id                            : chr [1:644] "A1" "A2" "A3" "A4" ...
 $ value                              : num [1:644] 202 195 244 211 178 222 204 187 232 207 ...
 $ plate_id                           : chr [1:644] "a" "a" "a" "a" ...
 $ time_hr                            : num [1:644] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank                           : logi [1:644] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis              : logi [1:644] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason                     : chr [1:644] NA NA NA NA ...
 $ family_id_group                    : chr [1:644] NA NA NA NA ...
 $ sample_id_group                    : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ treatment_group                    : chr [1:644] "control" "control" "oae" "oae" ...
 $ width_mm_measurement               : num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement              : num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement              : num [1:644] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement               : num [1:644] 7395 7340 6008 8749 8809 ...
 $ sample_key                         : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ value_t0                           : num [1:644] 202 195 244 211 178 222 204 187 232 207 ...
 $ fold_change                        : num [1:644] 1 1 1 1 1 1 1 1 1 1 ...
 $ trace_id                           : chr [1:644] "c-01" "c-25" "oae-03" "oae-27" ...
 $ mean_blank_rfu                     : num [1:644] 168 168 168 168 168 168 168 168 168 168 ...
 $ mean_blank_fc                      : num [1:644] 0.973 0.973 0.973 0.973 0.973 ...
 $ corrected_fc                       : num [1:644] 0.0275 0.0275 0.0275 0.0275 0.0275 ...
 $ metabolism_per_area_mm2_measurement: num [1:644] 3.72e-06 3.75e-06 4.58e-06 3.14e-06 3.12e-06 ...
if (has_fam) cat("## Mean metabolism by family\n\n")
if (nrow(metabolism_df) > 0 && has_fam) {

  metab_cols <- paste0("metabolism_per_", active_meas_cols)

  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    metab_summary <- metabolism_df %>%
      filter(!is.na(family_id_group), !exclude_from_analysis) %>%
      group_by(family_id_group, treatment_group, time_hr) %>%
      summarise(
        mean_val = mean(.data[[col]], na.rm = TRUE),
        se_val   = sd(.data[[col]], na.rm = TRUE) /
          sqrt(sum(!is.na(.data[[col]]))),
        .groups  = "drop"
      ) %>%
      mutate(group_var = if (has_trt)
        paste(family_id_group, treatment_group, sep = ".")
      else
        family_id_group)

    p_metab_mean <- ggplot(metab_summary,
        aes(x = time_hr, y = mean_val,
            colour = family_id_group,
            group = group_var)) +
      geom_ribbon(aes(ymin = mean_val - se_val,
                      ymax = mean_val + se_val,
                      fill = family_id_group),
                  alpha = 0.15, colour = NA) +
      geom_line(
        mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
        linewidth = 1) +
      geom_point(size = 2) +
      scale_colour_manual(values = fam_colours, name = "Family",
                          breaks = families) +
      scale_fill_manual(values = fam_colours, name = "Family",
                        breaks = families) +
      labs(x = "Time (h)",
           y = paste0(metabolism_y_label(col), " (± SE)")) +
      theme_classic(base_size = 13) +
      if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

    print(p_metab_mean)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_mean_", mc_label, ".png")),
      p_metab_mean, width = 8, height = 5)
  }
} else if (nrow(metabolism_df) > 0) {
  message("No family groups present: skipping mean metabolism by family.")
}
metab_cols <- paste0("metabolism_per_", active_meas_cols)
if (has_fam) cat("## Individual metabolism traces by family\n\n")
if (nrow(metabolism_df) > 0 && has_fam) {
  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    p_metab_by_family <- metabolism_df %>%
      filter(!is.na(family_id_group)) %>%
      ggplot(aes(x = time_hr, y = .data[[col]], group = trace_id,
                 colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
      geom_line(alpha = 0.6) +
      geom_point(size = 1.2, alpha = 0.7) +
      facet_wrap(~ family_id_group) +
      scale_colour_manual(
        values = if (has_trt) trt_colours else fam_colours,
        name   = if (has_trt) "Treatment" else "Family",
        breaks = if (has_trt) treatments else families) +
      labs(x = "Time (h)", y = metabolism_y_label(col)) +
      theme_classic(base_size = 12) +
      theme(strip.background = element_blank(),
            strip.text       = element_text(face = "bold"))

    print(p_metab_by_family)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_individual_", mc_label, "_by_family.png")),
      p_metab_by_family, width = 10, height = 5)
  }
}
if (has_trt) cat("## Individual metabolism traces by treatment\n\n")

9.1 Individual metabolism traces by treatment

if (nrow(metabolism_df) > 0 && has_trt) {
  colour_var  <- if (has_fam) "family_id_group" else "treatment_group"
  colour_vals <- if (has_fam) fam_colours else trt_colours
  colour_brks <- if (has_fam) families else treatments
  colour_name <- if (has_fam) "Family" else "Treatment"

  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    p_metab_by_treatment <- metabolism_df %>%
      ggplot(aes(x = time_hr, y = .data[[col]],
                 group = trace_id, colour = .data[[colour_var]])) +
      geom_line(alpha = 0.6) +
      geom_point(size = 1.2, alpha = 0.7) +
      facet_wrap(~ treatment_group) +
      scale_colour_manual(values = colour_vals, name = colour_name,
                          breaks = colour_brks) +
      labs(x = "Time (h)", y = metabolism_y_label(col)) +
      theme_classic(base_size = 12) +
      theme(strip.background = element_blank(),
            strip.text       = element_text(face = "bold"))

    print(p_metab_by_treatment)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_individual_", mc_label, "_by_treatment.png")),
      p_metab_by_treatment, width = 10, height = 5)
  }
}

10 Time-Series Statistical Analysis

Linear mixed effects models test the effect of experimental variables on metabolism over time. Individual (sample_id_group) is included as a random intercept to account for repeated measures across timepoints. Type III ANOVA with Satterthwaite’s approximation (lmerTest) assesses significance; post-hoc pairwise comparisons use estimated marginal means (emmeans, Tukey adjustment).

run_ts_stats <- function(df, value_col) {
  has_family    <- "family_id_group" %in% names(df) &&
    length(unique(na.omit(df$family_id_group))) > 1
  has_treatment <- "treatment_group" %in% names(df) &&
    length(unique(na.omit(df$treatment_group))) > 1

  if (!has_family && !has_treatment) return(NULL)

  df <- df %>%
    filter(is.finite(.data[[value_col]]), is.finite(time_hr)) %>%
    mutate(
      time_f     = factor(time_hr),
      individual = factor(trace_id)
    )

  if (nrow(df) == 0) return(NULL)

  if (has_family)    df <- df %>% mutate(family    = factor(family_id_group))
  if (has_treatment) df <- df %>% mutate(treatment = factor(treatment_group))

  if (has_family    && length(unique(na.omit(df$family)))    < 2) return(NULL)
  if (has_treatment && length(unique(na.omit(df$treatment))) < 2) return(NULL)

  fixed <- if (has_family && has_treatment) {
    paste0(value_col, " ~ time_f * family * treatment")
  } else if (has_family) {
    paste0(value_col, " ~ time_f * family")
  } else {
    paste0(value_col, " ~ time_f * treatment")
  }

  model <- lmer(
    as.formula(paste0(fixed, " + (1 | individual)")),
    data = df
  )

  anova_res <- anova(model, type = 3, ddf = "Satterthwaite")

  # Pairwise comparisons of group combinations at each timepoint
  emm_spec <- if (has_family && has_treatment) {
    ~ family * treatment | time_f
  } else if (has_family) {
    ~ family | time_f
  } else {
    ~ treatment | time_f
  }

  emm       <- emmeans(model, emm_spec)
  pairs_res <- as.data.frame(pairs(emm, adjust = "tukey"))

  # Main-effect marginal means (collapsed across time)
  emm_main <- if (has_family && has_treatment) {
    emmeans(model, ~ family * treatment)
  } else if (has_family) {
    emmeans(model, ~ family)
  } else {
    emmeans(model, ~ treatment)
  }

  pairs_main <- as.data.frame(pairs(emm_main, adjust = "tukey"))

  list(
    model         = model,
    anova         = anova_res,
    pairs_by_time = pairs_res,
    pairs_main    = pairs_main,
    has_family    = has_family,
    has_treatment = has_treatment
  )
}

ts_stats <- list()
if (nrow(metabolism_df) > 0) {
  for (mc in active_meas_cols) {
    col <- paste0("metabolism_per_", mc)
    if (col %in% names(metabolism_df))
      ts_stats[[col]] <- run_ts_stats(metabolism_df, col)
  }
}

10.1 Results

for (col in names(ts_stats)) {
  res <- ts_stats[[col]]
  if (is.null(res)) next

  cat("\n\n### Metric:", col, "\n\n")

  cat("**Type III ANOVA (Satterthwaite approximation):**\n\n")
  cat(knitr::kable(as.data.frame(res$anova), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")

  cat("**Marginal means – main effects (collapsed across time):**\n\n")
  cat(knitr::kable(as.data.frame(res$pairs_main), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")

  cat("**Pairwise comparisons by timepoint (Tukey):**\n\n")
  cat(knitr::kable(as.data.frame(res$pairs_by_time), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
}

10.1.1 Metric: metabolism_per_area_mm2_measurement

Type III ANOVA (Satterthwaite approximation):

Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
time_f 1e-04 0 6 540 60.5478 0.0000
treatment 0e+00 0 1 90 4.1251 0.0452
time_f:treatment 0e+00 0 6 540 2.4688 0.0230

Marginal means – main effects (collapsed across time):

contrast estimate SE df t.ratio p.value
control - oae -3e-04 1e-04 90 -2.031 0.0452

Pairwise comparisons by timepoint (Tukey):

contrast time_f estimate SE df t.ratio p.value
control - oae 0 0e+00 2e-04 168.2944 0.0045 0.9964
control - oae 0.5 -2e-04 2e-04 168.2944 -1.0383 0.3006
control - oae 1 -3e-04 2e-04 168.2944 -2.1690 0.0315
control - oae 1.5 -4e-04 2e-04 168.2944 -2.5397 0.0120
control - oae 2 -3e-04 2e-04 168.2944 -2.1645 0.0318
control - oae 2.5 -3e-04 2e-04 168.2944 -2.0021 0.0469
control - oae 3 -3e-04 2e-04 168.2944 -2.1752 0.0310

11 Area Under the Curve (AUC)

AUC computed per individual via the trapezoid rule across all timepoints. metabolism_per_* is the primary metric matching the paper; corrected_fc and raw_fluorescence are retained for reference.

compute_auc <- function(df, value_col, group_vars) {
  df %>%
    filter(is.finite(time_hr), is.finite(.data[[value_col]])) %>%
    group_by(across(all_of(group_vars))) %>%
    summarise(
      AUC          = trapezoid_auc(time_hr, .data[[value_col]]),
      n_timepoints = n(),
      .groups      = "drop"
    ) %>%
    filter(is.finite(AUC))
}

# Only include grouping columns that are actually present in the data
individual_vars <- intersect(
  c("trace_id", "family_id_group", "treatment_group"),
  names(metabolism_df)
)

auc_metab_list <- list()
if (nrow(metabolism_df) > 0) {
  for (mc in active_meas_cols) {
    col <- paste0("metabolism_per_", mc)
    if (col %in% names(metabolism_df)) {
      auc_metab_list[[col]] <-
        compute_auc(metabolism_df, col, individual_vars) %>%
        mutate(metric = col)
    }
  }
}

auc_all <- bind_rows(auc_metab_list)

str(auc_all)
tibble [92 × 6] (S3: tbl_df/tbl/data.frame)
 $ trace_id       : chr [1:92] "c-01" "c-02" "c-03" "c-04" ...
 $ family_id_group: chr [1:92] NA NA NA NA ...
 $ treatment_group: chr [1:92] "control" "control" "control" "control" ...
 $ AUC            : num [1:92] 0.000988 0.000663 0.001028 0.001147 0.001497 ...
 $ n_timepoints   : int [1:92] 7 7 7 7 7 7 7 7 7 7 ...
 $ metric         : chr [1:92] "metabolism_per_area_mm2_measurement" "metabolism_per_area_mm2_measurement" "metabolism_per_area_mm2_measurement" "metabolism_per_area_mm2_measurement" ...

11.1 AUC summary tables

sum_vars <- intersect(
  c("metric", "family_id_group", "treatment_group"),
  names(auc_all)
)
auc_summary <- auc_all %>%
  group_by(across(all_of(sum_vars))) %>%
  summarise(
    n      = n(),
    mean   = mean(AUC, na.rm = TRUE),
    sd     = sd(AUC, na.rm = TRUE),
    se     = sd / sqrt(n),
    median = median(AUC, na.rm = TRUE),
    .groups = "drop"
  )

print(auc_summary)
# A tibble: 2 × 8
  metric   family_id_group treatment_group     n    mean      sd      se  median
  <chr>    <chr>           <chr>           <int>   <dbl>   <dbl>   <dbl>   <dbl>
1 metabol… <NA>            control            46 0.00129 0.00155 2.29e-4 7.96e-4
2 metabol… <NA>            oae                46 0.00215 0.00233 3.43e-4 9.93e-4

12 Statistical Analysis

Each individual oyster (sample_id_group) is the observational unit. The model is built from whichever grouping factors are present: both family and treatment (with interaction) when both exist, or a one-way model when only one factor is available.

run_auc_stats <- function(auc_df) {
  empty <- tibble()

  has_family    <- "family_id_group" %in% names(auc_df) &&
    length(unique(na.omit(auc_df$family_id_group))) > 1
  has_treatment <- "treatment_group" %in% names(auc_df) &&
    length(unique(na.omit(auc_df$treatment_group))) > 1

  if (!has_family && !has_treatment) {
    return(list(model = NULL, anova = NULL,
                pairs_full = empty, pairs_family = empty,
                pairs_trt = empty,
                has_family = FALSE, has_treatment = FALSE))
  }

  if (has_family)    auc_df <- auc_df %>% mutate(family    = factor(family_id_group))
  if (has_treatment) auc_df <- auc_df %>% mutate(treatment = factor(treatment_group))

  formula_str <- if (has_family && has_treatment) {
    "AUC ~ family * treatment"
  } else if (has_family) {
    "AUC ~ family"
  } else {
    "AUC ~ treatment"
  }
  model     <- lm(as.formula(formula_str), data = auc_df)
  anova_res <- anova(model)

  if (has_family && has_treatment) {
    pairs_full   <- as.data.frame(pairs(emmeans(model, ~ family * treatment),
                                        adjust = "tukey"))
    pairs_family <- as.data.frame(pairs(emmeans(model, ~ family),
                                        adjust = "tukey"))
    pairs_trt    <- as.data.frame(pairs(emmeans(model, ~ treatment),
                                        adjust = "tukey"))
  } else if (has_family) {
    pairs_family <- as.data.frame(pairs(emmeans(model, ~ family),
                                        adjust = "tukey"))
    pairs_full   <- pairs_family
    pairs_trt    <- empty
  } else {
    pairs_trt    <- as.data.frame(pairs(emmeans(model, ~ treatment),
                                        adjust = "tukey"))
    pairs_full   <- pairs_trt
    pairs_family <- empty
  }

  list(
    model         = model,
    anova         = anova_res,
    pairs_full    = pairs_full,
    pairs_family  = pairs_family,
    pairs_trt     = pairs_trt,
    has_family    = has_family,
    has_treatment = has_treatment
  )
}

metrics_to_test <- unique(auc_all$metric)
stats_results   <- map(
  set_names(metrics_to_test),
  ~ run_auc_stats(auc_all %>% filter(metric == .x))
)

12.1 Results by metric

for (met in metrics_to_test) {
  stats <- stats_results[[met]]
  cat("\n\n### Metric:", met, "\n\n")
  cat("**ANOVA:**\n\n")
  cat(knitr::kable(as.data.frame(stats$anova), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  if (stats$has_family && stats$has_treatment) {
    cat("**Pairwise: family × treatment (Tukey):**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_full), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
    cat("**Pairwise: family main effect:**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_family), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
    cat("**Pairwise: treatment main effect:**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_trt), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  } else if (stats$has_family) {
    cat("**Pairwise: family (Tukey):**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_family), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  } else if (stats$has_treatment) {
    cat("**Pairwise: treatment (Tukey):**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_trt), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  }
}

12.1.1 Metric: metabolism_per_area_mm2_measurement

ANOVA:

Df Sum Sq Mean Sq F value Pr(>F)
treatment 1 0e+00 0 4.3289 0.0403
Residuals 90 4e-04 0 NA NA

Pairwise: treatment (Tukey):

contrast estimate SE df t.ratio p.value
control - oae -9e-04 4e-04 90 -2.0806 0.0403

13 AUC Box Plots with Statistical Annotations

Significance labels: *** p < 0.001, ** p < 0.01, * p < 0.05. Brackets are drawn only for significant pairs (p < 0.05). Plots are generated for whichever grouping factors are present: treatment-only, family-only, all-groups, within-family, and within-treatment.

sig_label <- function(p) {
  case_when(p < 0.001 ~ "***", p < 0.01 ~ "**", p < 0.05 ~ "*",
            TRUE ~ "ns")
}

# Add significance brackets to an existing ggplot.
# pairs_df   : data frame with $contrast and $p.value columns
# group_levels: ordered character vector matching x-axis factor levels
# y_vals     : numeric vector of AUC values used to set bracket heights
add_sig_brackets <- function(p, pairs_df, group_levels, y_vals) {
  sig_pairs <- pairs_df %>%
    mutate(label = sig_label(p.value)) %>%
    filter(label != "ns")
  if (nrow(sig_pairs) == 0) return(p)

  y_max   <- max(y_vals, na.rm = TRUE)
  y_range <- diff(range(y_vals, na.rm = TRUE))
  step    <- y_range * 0.12

  # emmeans prefixes factor levels with the variable name (e.g. "family7");
  # resolve back to the bare level used on the x-axis.
  resolve_level <- function(x, lvls) {
    x <- trimws(x)
    if (x %in% lvls) return(x)
    lvls_sorted <- lvls[order(nchar(as.character(lvls)), decreasing = TRUE)]
    for (lvl in lvls_sorted) {
      lvl_str <- as.character(lvl)
      if (endsWith(x, lvl_str) && nchar(x) > nchar(lvl_str)) return(lvl_str)
    }
    NA_character_
  }

  for (i in seq_len(nrow(sig_pairs))) {
    parts <- str_split(as.character(sig_pairs$contrast[i]), " - ", 2)[[1]]
    g1 <- resolve_level(parts[1], group_levels)
    g2 <- resolve_level(parts[2], group_levels)
    x1 <- match(g1, group_levels)
    x2 <- match(g2, group_levels)
    if (is.na(x1) || is.na(x2)) next
    bar_y <- y_max + i * step
    p <- p +
      annotate("segment", x = x1, xend = x2,
               y = bar_y, yend = bar_y,
               colour = "black", linewidth = 0.6) +
      annotate("segment", x = x1, xend = x1,
               y = bar_y, yend = bar_y - step * 0.3,
               colour = "black", linewidth = 0.6) +
      annotate("segment", x = x2, xend = x2,
               y = bar_y, yend = bar_y - step * 0.3,
               colour = "black", linewidth = 0.6) +
      annotate("text", x = (x1 + x2) / 2,
               y = bar_y + step * 0.15,
               label = sig_pairs$label[i], size = 4.5)
  }
  p
}
for (met in metrics_to_test) {
  df      <- auc_all %>% filter(metric == met)
  stats   <- stats_results[[met]]
  y_lab   <- auc_y_label(met)
  has_fam <- stats$has_family
  has_trt <- stats$has_treatment

  # ── Treatment main effect (x = treatment, tick = treatment name) ───────
  if (has_trt) {
    df_p <- df %>%
      mutate(x = factor(treatment_group, levels = sort(unique(treatment_group))))
    grps <- levels(df_p$x)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = trt_colours[grps], guide = "none") +
      labs(x = "Treatment", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, stats$pairs_trt, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_treatment_", met, ".png")),
           p, width = 5, height = 5)
  }

  # ── Family main effect (x = family, tick = family name) ───────────────
  if (has_fam) {
    df_p <- df %>%
      mutate(x = factor(family_id_group,
                        levels = str_sort(unique(family_id_group), numeric = TRUE)))
    grps <- levels(df_p$x)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = fam_colours[grps], guide = "none") +
      labs(x = "Family", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, stats$pairs_family, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_family_", met, ".png")),
           p, width = 5, height = 5)
  }

  # Remaining plots require both factors
  if (!has_fam || !has_trt) next

  # ── All family:treatment groups (x = family:treatment) ─────────────────
  # emmeans contrasts use spaces; convert to colon to match tick labels
  pairs_fc <- stats$pairs_full %>%
    mutate(contrast = str_replace_all(
      contrast,
      "([a-z]+) ([a-z]+)( - )([a-z]+) ([a-z]+)",
      "\\1:\\2\\3\\4:\\5"
    ))
  df_p <- df %>%
    mutate(x = factor(
      paste(family_id_group, treatment_group, sep = ":"),
      levels = str_sort(unique(paste(family_id_group, treatment_group, sep = ":")),
                        numeric = TRUE)
    ))
  grps     <- levels(df_p$x)
  fill_map <- setNames(make_palette(length(grps)), grps)
  p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
    geom_boxplot(alpha = 0.6, outlier.shape = NA) +
    geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
    scale_fill_manual(values = fill_map, guide = "none") +
    labs(x = "Family : Treatment", y = y_lab) +
    theme_classic(base_size = 13) +
    theme(axis.text.x = element_text(angle = 20, hjust = 1))
  p <- add_sig_brackets(p, pairs_fc, grps, df_p$AUC)
  print(p)
  ggsave(file.path(fig_dir, paste0("auc_all_groups_", met, ".png")),
         p, width = 6, height = 5)

  # ── Within each family: treatment comparison (x = family:treatment) ────
  # Tick labels are family:treatment so these plots are visually distinct
  # from the treatment main-effect plot above.
  for (fam in str_sort(unique(df$family_id_group), numeric = TRUE)) {
    df_p <- df %>%
      filter(family_id_group == fam) %>%
      mutate(x = factor(
        paste(family_id_group, treatment_group, sep = ":"),
        levels = str_sort(unique(paste(family_id_group, treatment_group, sep = ":")),
                          numeric = TRUE)
      ))
    grps     <- levels(df_p$x)
    pairs_sub <- pairs_fc %>%
      filter(str_count(contrast, paste0(fam, ":")) == 2)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = fill_map[grps], guide = "none") +
      labs(x = "Family : Treatment", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, pairs_sub, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_", fam, "_trt_", met, ".png")),
           p, width = 5, height = 5)
  }

  # ── Within each treatment: family comparison (x = family:treatment) ────
  # Tick labels are family:treatment so these plots are visually distinct
  # from the family main-effect plot above.
  for (trt in sort(unique(df$treatment_group))) {
    df_p <- df %>%
      filter(treatment_group == trt) %>%
      mutate(x = factor(
        paste(family_id_group, treatment_group, sep = ":"),
        levels = str_sort(unique(paste(family_id_group, treatment_group, sep = ":")),
                          numeric = TRUE)
      ))
    grps      <- levels(df_p$x)
    pairs_sub <- pairs_fc %>%
      filter(str_count(contrast, paste0(":", trt)) == 2)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = fill_map[grps], guide = "none") +
      labs(x = "Family : Treatment", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, pairs_sub, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_", trt, "_fam_", met, ".png")),
           p, width = 5, height = 5)
  }
}

14 Save Output Data

write_csv(auc_all,      file.path(out_dir, "auc_all_metrics.csv"))
write_csv(auc_summary,  file.path(out_dir, "auc_summary.csv"))

if (nrow(metabolism_df) > 0)
  write_csv(metabolism_df,
            file.path(out_dir, "metabolism.csv"))

stats_compiled <- map_dfr(metrics_to_test, function(met) {
  bind_rows(
    stats_results[[met]]$pairs_full %>%
      mutate(comparison = "family:treatment"),
    stats_results[[met]]$pairs_family %>%
      mutate(comparison = "family"),
    stats_results[[met]]$pairs_trt %>%
      mutate(comparison = "treatment")
  ) %>% mutate(metric = met)
})

write_csv(stats_compiled,
          file.path(out_dir, "pairwise_stats.csv"))

message("Output files written to: ", out_dir)
]]> 2026 resazurin Magallana gigas Crassostrea gigas Pacific oyster heat stress OAE https://robertslab.github.io/sams-notebook/posts/2026/2026-08-31-Resazurin-Assays---USDA-M.gigas-Juveniles-OAE-Preconditioning-Effect-on-36C-Heat-Stress-Response/ Mon, 31 Aug 2026 07:00:00 GMT Citrate Synthase Analysis - SORMI June 2026 M.gigas Family 5 vs Family 7 Temperature Response Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-25-Citrate-Synthase-Analysis---SORMI-June-2026-M.gigas-Family-5-vs-Family-7-Temperature-Response/

INTRO

Cross-plate comparison of citrate synthase (CS) activity across two families (F05, F07) at two temperature exposures (ambient and 36°C), drawing on the finished per-plate results from three prior notebook entries: F05 ambient re-assay, 20260824, F05 36°C, 20260814, and F07 ambient + 36°C, 20260824 (Notebook entries). The 2026-08-11 F05 ambient plate (Notebook entry) is deliberately excluded — it required repeating for poor data quality — and its 20260824 re-assay is used in its place. All four family × temperature groups carry n = 8 individuals, a fully balanced 2 × 2 design across 32 individuals.

The question this analysis asks is whether CS activity responds to 36°C exposure, and whether that response differs between families (a family × temperature interaction).

Background luminescence is not subtracted from any activity value in this document. One retained plate (F07 ambient + 36°C) has no usable background estimate, so to treat every plate identically, the background correction is removed from all three source analyses before comparison. See the BACKGROUND REMOVAL section below for how small this correction was where it could be measured.

All data/code/analysis is available here:

MATERIALS & METHODS

This is a downstream statistical analysis, not a new plate assay — it takes the finished citrate_synthase_activity_results.csv output of the three per-plate analyses linked above as its input. All individuals were originally assayed with the Citrate Synthase Assay Kit (ABCAM; cat: ab239712); see the linked per-plate entries for assay methods, and the homogenization and protein quantification entries for upstream methods.

Individuals are treated as unpaired across temperature — an animal is assayed at one exposure only. Statistics are computed on log10(activity): a two-way ANOVA (family × temperature) on a balanced design (n = 8/cell), followed by Tukey HSD pairwise contrasts. The F05 ambient-vs-36°C comparison spans two plates run ten days apart (20260814, 20260824); the F07 comparison is made within a single plate (20260824) — cross-plate comparability is checked against standard-curve and positive-control calibration values in the PLATE PROVENANCE section below.

Analysis

Data were analyzed in R. The Rmd file is here (GitHub):

Knitted markdown is below.

Data files

Output tables and figures from this analysis are here:

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/outputs/Gen5-20260825-mgig-sormi-citrate_synthase-F05-F07-temperature-comparison

AI Use Level 4: Substantial AI contribution

RESULTS

The temperature response differs between the two families. A two-way ANOVA on log10(activity) finds a significant family × temperature interaction (p = 1e-05), accounting for 23.0% of total variance:

  • F05 loses CS activity at 36°C — mean activity falls from 114.2 to 40.6 mU/mg protein, a 0.36-fold change (-64%), significant after Tukey adjustment (p = 2e-07).

  • F07 shows essentially no temperature response — mean activity moves from 129.5 to 129.0 mU/mg protein (1.00-fold, p = 1); its 95% CI at 36°C overlaps its ambient CI substantially.

  • The two families are indistinguishable at ambient temperature (1.10-fold, p = 0.9) — they differ only after heat exposure, which is what makes this an interaction rather than a baseline family difference.

  • Removing the background correction does not drive any of this. Across all 32 individuals the correction was worth 0.0-11.3% of activity (median 0.8%), and both families’ ambient groups shift in the same direction.

1 BACKGROUND

Cross-plate comparison of citrate synthase (CS) activity in ctenidia of Magallana gigas (Pacific oyster) from two families (F05, F07) at two temperature exposures (ambient and 36 °C), assayed with the Abcam Citrate Synthase Assay Kit (ab239712), v4a.

The question here is not how to compute activity – that was done, plate by plate, in the per-plate analyses listed below. This document takes those finished per-plate results as its input and asks:

  1. Does CS activity respond to 36 °C exposure?
  2. Does that temperature response differ between families? – i.e. is there a family × temperature interaction?

All four family × temperature groups carry n = 8 individuals, giving a fully balanced 2 × 2 design (32 individuals total).

1.1 Inputs: which plates are used

This analysis reads the citrate_synthase_activity_results.csv output of three per-plate analyses:

Family Temperature Source analysis Plate date
F05 ambient Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient 2026-08-24
F05 36 °C Gen5-20260814-mgig-citrate_synthase-F05-36C 2026-08-14
F07 ambient + 36 °C Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C 2026-08-24

The 2026-08-11 F05 ambient plate is deliberately excluded (note 1).

1.2 Notes

  1. The 2026-08-11 F05-ambient plate is excluded from this analysis. That plate required repeating because of poor data quality across both its samples and its standard curve (standard-curve R2 of 0.9678 after excluding 8 of 18 standard wells as outliers, 4 of 8 samples over the 15% technical-CV threshold, and one sample with all three replicates baseline-compromised). Those same F05 ambient homogenates were re-assayed on 2026-08-24, and the 2026-08-24 re-assay is the F05 ambient data used here.
  2. Background luminescence is NOT subtracted from any activity value in this document. One retained plate (Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C) has no usable background estimate – Reaction Mix was added to all three of its pooled-background wells instead of Background Control Mix, so its background rate is unestimable and its per-plate analysis already fell back to a correction of 0. Rather than compare background-corrected values on two plates against uncorrected values on a third, the background correction is removed from all plates so every value is treated identically. This is a small change: where background was measurable it never exceeded ~10% of the raw rate (quantified in BACKGROUND REMOVAL).
  3. Activity is normalized to total extracted protein (mU per mg protein), carried through from the per-plate analyses, which measured protein directly by BCA/Bradford-style assay on 2026-08-13. Tissue weight is not used.
  4. The F05 temperature contrast spans two plates and two days (36 °C on 2026-08-14, ambient on 2026-08-24), whereas the F07 temperature contrast sits within a single plate (2026-08-24). Plate and instrument state are compared across plates in PLATE PROVENANCE.
  5. Individuals are treated as unpaired across temperature. An animal is assayed at one exposure only; F05_01_ambient and F05_01_36C are different individuals from the same family, so no paired/repeated-measures structure is used.
  6. Statistics are computed on log10(activity). Activity is a ratio-scale rate spanning roughly an order of magnitude across groups; the log scale makes the family × temperature interaction a test of multiplicative (fold-change) response, which is the biologically meaningful comparison for an enzyme rate, and it stabilizes the group-to-group spread in variance. Because the design is balanced, sequential (Type I) sums of squares from aov() are identical to Type II and Type III, so no special contrast handling is needed.
library(knitr)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)
library(ggplot2)
library(tibble)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

1.3 Analysis parameters

analysis_params <- list(
  subtract_background = FALSE,  # note 2: background is NOT subtracted anywhere
  alpha               = 0.05,   # significance level for ANOVA / Tukey
  cv_threshold_pct    = 15      # technical-CV threshold used by the per-plate analyses
)

cat("--- analysis_params: settings governing this comparison ---\n")
str(analysis_params)
--- analysis_params: settings governing this comparison ---
List of 3
 $ subtract_background: logi FALSE
 $ alpha              : num 0.05
 $ cv_threshold_pct   : num 15

1.4 Input paths

outputs_root <- "../outputs"

# Each retained per-plate analysis and the family/temperature it contributes.
plate_inputs <- tribble(
  ~plate_dir,                                                       ~contributes,
  "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient",          "F05 ambient",
  "Gen5-20260814-mgig-citrate_synthase-F05-36C",                    "F05 36C",
  "Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C",  "F07 ambient + 36C"
) %>%
  mutate(results_csv = file.path(outputs_root, plate_dir,
                                 "citrate_synthase_activity_results.csv"))

output_dir <- file.path(
  outputs_root,
  "Gen5-20260825-mgig-sormi-citrate_synthase-F05-F07-temperature-comparison")
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- plate_inputs: per-plate results files feeding this analysis ---\n")
print(as.data.frame(plate_inputs[, c("plate_dir", "contributes")]))
cat("\n--- all input files exist? ---\n")
print(setNames(file.exists(plate_inputs$results_csv), plate_inputs$plate_dir))
cat("\n--- output_dir ---\n")
str(output_dir)
--- plate_inputs: per-plate results files feeding this analysis ---
                                                      plate_dir
1         Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient
2                   Gen5-20260814-mgig-citrate_synthase-F05-36C
3 Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C
        contributes
1       F05 ambient
2           F05 36C
3 F07 ambient + 36C

--- all input files exist? ---
        Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient 
                                                         TRUE 
                  Gen5-20260814-mgig-citrate_synthase-F05-36C 
                                                         TRUE 
Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C 
                                                         TRUE 

--- output_dir ---
 chr "../outputs/Gen5-20260825-mgig-sormi-citrate_synthase-F05-F07-temperature-comparison"

1.5 Plate provenance

Per-plate calibration and instrument-state values, transcribed from the three source analyses. These are the anchors that make cross-plate comparison defensible: if the GSH standard curve and the CS positive control agree across plates, then a rate measured on one plate means the same thing as a rate measured on another.

plate_provenance <- tribble(
  ~plate_dir,                                                      ~plate_date,  ~std_slope_A412_per_nmol, ~std_r2, ~pos_control_mA412_min, ~pos_control_cv_pct, ~background_usable,
  "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient",         "2026-08-24", 0.03234,                  0.9983,  2.700,                  22.8,                TRUE,
  "Gen5-20260814-mgig-citrate_synthase-F05-36C",                   "2026-08-14", 0.03075,                  0.9985,  2.933,                  3.9,                 TRUE,
  "Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C", "2026-08-24", 0.03140,                  0.9992,  2.933,                  2.6,                 FALSE
)

cat("--- plate_provenance: calibration + instrument state per plate ---\n")
str(as.data.frame(plate_provenance))
--- plate_provenance: calibration + instrument state per plate ---
'data.frame':   3 obs. of  7 variables:
 $ plate_dir              : chr  "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient" "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C"
 $ plate_date             : chr  "2026-08-24" "2026-08-14" "2026-08-24"
 $ std_slope_A412_per_nmol: num  0.0323 0.0307 0.0314
 $ std_r2                 : num  0.998 0.999 0.999
 $ pos_control_mA412_min  : num  2.7 2.93 2.93
 $ pos_control_cv_pct     : num  22.8 3.9 2.6
 $ background_usable      : logi  TRUE TRUE FALSE
plate_provenance %>%
  transmute(
    Plate            = plate_dir,
    Date             = plate_date,
    `Std slope (A412/nmol)` = std_slope_A412_per_nmol,
    `Std curve R^2`  = std_r2,
    `Pos ctrl (mA412/min)` = pos_control_mA412_min,
    `Pos ctrl CV (%)` = pos_control_cv_pct,
    `Background estimable` = ifelse(background_usable, "yes", "no (note 2)")
  ) %>%
  kable(caption = "Calibration and instrument state for each retained plate.")
Plate Date Std slope (A412/nmol) Std curve R^2 Pos ctrl (mA412/min) Pos ctrl CV (%) Background estimable
Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient 2026-08-24 0.03234 0.9983 2.700 22.8 yes
Gen5-20260814-mgig-citrate_synthase-F05-36C 2026-08-14 0.03075 0.9985 2.933 3.9 yes
Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C 2026-08-24 0.03140 0.9992 2.933 2.6 no (note 2)

Calibration and instrument state for each retained plate.

prov_spread <- plate_provenance %>%
  summarise(
    slope_min   = min(std_slope_A412_per_nmol),
    slope_max   = max(std_slope_A412_per_nmol),
    slope_cv    = 100 * sd(std_slope_A412_per_nmol) / mean(std_slope_A412_per_nmol),
    pos_min     = min(pos_control_mA412_min),
    pos_max     = max(pos_control_mA412_min),
    pos_cv      = 100 * sd(pos_control_mA412_min) / mean(pos_control_mA412_min),
    min_std_r2  = min(std_r2)
  )

cat(sprintf(
  "Standard-curve slope across plates: %.5f-%.5f A412/nmol (CV %.1f%%)\n",
  prov_spread$slope_min, prov_spread$slope_max, prov_spread$slope_cv))
cat(sprintf(
  "Positive control across plates:     %.3f-%.3f mA412/min (CV %.1f%%)\n",
  prov_spread$pos_min, prov_spread$pos_max, prov_spread$pos_cv))
cat(sprintf("Worst standard-curve R^2:           %.4f\n", prov_spread$min_std_r2))
Standard-curve slope across plates: 0.03075-0.03234 A412/nmol (CV 2.5%)
Positive control across plates:     2.700-2.933 mA412/min (CV 4.7%)
Worst standard-curve R^2:           0.9983

2 LOAD PER-PLATE RESULTS

2.1 Read the results CSVs

read_plate_results <- function(results_csv, plate_dir) {
  read.csv(results_csv, check.names = FALSE, stringsAsFactors = FALSE) %>%
    mutate(plate_dir = plate_dir)
}

plate_results_raw <- Map(read_plate_results,
                         plate_inputs$results_csv,
                         plate_inputs$plate_dir) %>%
  bind_rows()

cat("--- plate_results_raw: all per-plate result rows, stacked ---\n")
str(as.data.frame(plate_results_raw))
--- plate_results_raw: all per-plate result rows, stacked ---
'data.frame':   32 obs. of  18 variables:
 $ Sample                    : chr  "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ Family                    : chr  "F05" "F05" "F05" "F05" ...
 $ Individual                : int  1 2 3 4 5 6 7 8 1 2 ...
 $ Temperature               : chr  "ambient" "ambient" "ambient" "ambient" ...
 $ Protein conc (ug/mL)      : num  876 1039 1900 1129 712 ...
 $ Total protein (mg)        : num  0.307 0.364 0.665 0.395 0.249 0.781 0.591 0.946 0.334 0.191 ...
 $ Reps used                 : chr  "3/3" "3/3" "3/3" "3/3" ...
 $ CV all reps (%)           : num  4.6 4.1 2.2 2.5 1.9 1.9 1.3 0.8 3.3 2.4 ...
 $ CV used reps (%)          : num  4.6 4.1 2.2 2.5 1.9 1.9 1.3 0.8 3.3 2.4 ...
 $ Rate (mA412/min)          : num  7.6 8.5 13.48 10.18 5.62 ...
 $ BG rate (mA412/min)       : num  0.233 0.233 0.233 0.233 0.233 0.233 0.233 0.233 0.167 0.15 ...
 $ Corrected rate (mA412/min): num  7.37 8.27 13.25 9.95 5.38 ...
 $ Activity (mU/uL)          : num  0.1139 0.1278 0.2048 0.1538 0.0832 ...
 $ Activity (mU/mg protein)  : num  130 123 108 136 117 ...
 $ CV flag                   : chr  "pass" "pass" "pass" "pass" ...
 $ Elevated baseline reps    : chr  "0/3" "0/3" "0/3" "0/3" ...
 $ Interpretation            : chr  "usable" "usable" "usable" "usable" ...
 $ plate_dir                 : chr  "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient" "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient" "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient" "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient" ...
cat("Rows read per plate:\n")
print(table(plate_results_raw$plate_dir))
cat("\nColumns present in every input file:\n")
print(names(plate_results_raw))
Rows read per plate:

                  Gen5-20260814-mgig-citrate_synthase-F05-36C 
                                                            8 
        Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient 
                                                            8 
Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C 
                                                           16 

Columns present in every input file:
 [1] "Sample"                     "Family"                    
 [3] "Individual"                 "Temperature"               
 [5] "Protein conc (ug/mL)"       "Total protein (mg)"        
 [7] "Reps used"                  "CV all reps (%)"           
 [9] "CV used reps (%)"           "Rate (mA412/min)"          
[11] "BG rate (mA412/min)"        "Corrected rate (mA412/min)"
[13] "Activity (mU/uL)"           "Activity (mU/mg protein)"  
[15] "CV flag"                    "Elevated baseline reps"    
[17] "Interpretation"             "plate_dir"

2.2 Keep only the columns this analysis needs

The per-plate results tables carry QC columns alongside the activity values. This analysis needs the identity of each individual, its protein normalization, both the raw and background-corrected rate (to undo the background subtraction – note 2), the published activity, and the QC verdict.

cs_all <- plate_results_raw %>%
  transmute(
    sample_id        = Sample,
    family           = Family,
    temperature      = Temperature,
    individual       = sprintf("%02d", as.integer(Individual)),
    plate_dir        = plate_dir,
    protein_mg       = `Total protein (mg)`,
    reps_used        = `Reps used`,
    cv_used_pct      = `CV used reps (%)`,
    rate_raw         = `Rate (mA412/min)`,
    rate_bg          = `BG rate (mA412/min)`,
    rate_corrected   = `Corrected rate (mA412/min)`,
    activity_published = `Activity (mU/mg protein)`,
    cv_flag          = `CV flag`,
    interpretation   = Interpretation
  ) %>%
  left_join(plate_provenance %>% select(plate_dir, plate_date, std_slope_A412_per_nmol),
            by = "plate_dir") %>%
  arrange(family, temperature, individual)

cat("--- cs_all: one row per individual, columns needed downstream ---\n")
str(as.data.frame(cs_all))
--- cs_all: one row per individual, columns needed downstream ---
'data.frame':   32 obs. of  16 variables:
 $ sample_id              : chr  "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ family                 : chr  "F05" "F05" "F05" "F05" ...
 $ temperature            : chr  "36C" "36C" "36C" "36C" ...
 $ individual             : chr  "01" "02" "03" "04" ...
 $ plate_dir              : chr  "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260814-mgig-citrate_synthase-F05-36C" ...
 $ protein_mg             : num  0.334 0.191 0.234 0.372 0.583 0.443 0.389 0.475 0.307 0.364 ...
 $ reps_used              : chr  "3/3" "3/3" "3/3" "3/3" ...
 $ cv_used_pct            : num  3.3 2.4 3.5 2.3 1.2 2.4 1.2 7.8 4.6 4.1 ...
 $ rate_raw               : num  2.33 2.37 2.45 2.47 2.33 2.37 2.37 2.43 7.6 8.5 ...
 $ rate_bg                : num  0.167 0.15 0.15 0.183 0.183 0.233 0.167 0.183 0.233 0.233 ...
 $ rate_corrected         : num  2.17 2.22 2.3 2.28 2.15 2.13 2.2 2.25 7.37 8.27 ...
 $ activity_published     : num  37 66 56 35 21 ...
 $ cv_flag                : chr  "pass" "pass" "pass" "pass" ...
 $ interpretation         : chr  "usable" "usable" "usable" "usable" ...
 $ plate_date             : chr  "2026-08-14" "2026-08-14" "2026-08-14" "2026-08-14" ...
 $ std_slope_A412_per_nmol: num  0.0307 0.0307 0.0307 0.0307 0.0307 ...

2.3 Verify the design is complete and balanced

design_counts <- cs_all %>%
  count(family, temperature, name = "n_individuals")

cat("--- design_counts: individuals per family x temperature cell ---\n")
print(as.data.frame(design_counts))

cat(sprintf("\nTotal individuals: %d\n", nrow(cs_all)))
cat(sprintf("Duplicated sample_ids: %d\n", sum(duplicated(cs_all$sample_id))))
cat(sprintf("Rows with missing protein_mg: %d\n", sum(is.na(cs_all$protein_mg))))
cat(sprintf("Rows with missing raw rate: %d\n", sum(is.na(cs_all$rate_raw))))
cat(sprintf("Design balanced (all cells equal n): %s\n",
            length(unique(design_counts$n_individuals)) == 1))
--- design_counts: individuals per family x temperature cell ---
  family temperature n_individuals
1    F05         36C             8
2    F05     ambient             8
3    F07         36C             8
4    F07     ambient             8

Total individuals: 32
Duplicated sample_ids: 0
Rows with missing protein_mg: 0
Rows with missing raw rate: 0
Design balanced (all cells equal n): TRUE

3 BACKGROUND REMOVAL

3.1 Why the correction is removed

Per note 2, the F07 plate has no estimable background, so its published “corrected” rate already equals its raw rate. The two F05 plates do carry a background subtraction. Comparing corrected F05 values against uncorrected F07 values would put a systematic, plate-dependent offset directly into the family contrast – so instead the correction is stripped from every plate.

3.2 How much does the correction actually matter?

bg_impact <- cs_all %>%
  mutate(bg_pct_of_raw = 100 * rate_bg / rate_raw) %>%
  group_by(plate_dir) %>%
  summarise(
    n              = n(),
    bg_rate_min    = min(rate_bg),
    bg_rate_max    = max(rate_bg),
    bg_pct_min     = min(bg_pct_of_raw),
    bg_pct_max     = max(bg_pct_of_raw),
    .groups = "drop"
  )

cat("--- bg_impact: size of the background correction, per plate ---\n")
print(as.data.frame(bg_impact))
--- bg_impact: size of the background correction, per plate ---
                                                      plate_dir  n bg_rate_min
1                   Gen5-20260814-mgig-citrate_synthase-F05-36C  8       0.150
2         Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient  8       0.233
3 Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C 16       0.000
  bg_rate_max bg_pct_min bg_pct_max
1       0.233   6.122449   9.831224
2       0.233   1.581806   4.145907
3       0.000   0.000000   0.000000
bg_impact %>%
  transmute(
    Plate = plate_dir,
    n     = n,
    `BG rate (mA412/min)` = sprintf("%.3f - %.3f", bg_rate_min, bg_rate_max),
    `BG as % of raw rate` = sprintf("%.1f - %.1f", bg_pct_min, bg_pct_max)
  ) %>%
  kable(caption = paste("Magnitude of the background correction being removed.",
                        "The F07 plate has no estimable background (note 2),",
                        "so its correction is already zero."))
Plate n BG rate (mA412/min) BG as % of raw rate
Gen5-20260814-mgig-citrate_synthase-F05-36C 8 0.150 - 0.233 6.1 - 9.8
Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient 8 0.233 - 0.233 1.6 - 4.1
Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C 16 0.000 - 0.000 0.0 - 0.0

Magnitude of the background correction being removed. The F07 plate has no estimable background (note 2), so its correction is already zero.

3.3 Recompute activity without background subtraction

Activity is linear in the background-corrected rate: every other term in the Abcam calculation (standard-curve slope, sample volume, dilution factor, homogenate volume, total protein) is a per-sample constant. So the uncorrected activity is recovered exactly by rescaling the published activity by rate_raw / rate_corrected, with no need to re-derive anything from the raw kinetic traces.

stopifnot(!analysis_params$subtract_background)

cs_activity <- cs_all %>%
  mutate(
    # Scale factor undoing the background subtraction; exactly 1 where the
    # per-plate analysis already applied no correction (F07 plate).
    bg_undo_factor  = rate_raw / rate_corrected,
    activity_nobg   = activity_published * bg_undo_factor,
    log10_activity  = log10(activity_nobg)
  )

cat("--- cs_activity: activity with background subtraction removed ---\n")
str(as.data.frame(cs_activity))
--- cs_activity: activity with background subtraction removed ---
'data.frame':   32 obs. of  19 variables:
 $ sample_id              : chr  "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ family                 : chr  "F05" "F05" "F05" "F05" ...
 $ temperature            : chr  "36C" "36C" "36C" "36C" ...
 $ individual             : chr  "01" "02" "03" "04" ...
 $ plate_dir              : chr  "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260814-mgig-citrate_synthase-F05-36C" "Gen5-20260814-mgig-citrate_synthase-F05-36C" ...
 $ protein_mg             : num  0.334 0.191 0.234 0.372 0.583 0.443 0.389 0.475 0.307 0.364 ...
 $ reps_used              : chr  "3/3" "3/3" "3/3" "3/3" ...
 $ cv_used_pct            : num  3.3 2.4 3.5 2.3 1.2 2.4 1.2 7.8 4.6 4.1 ...
 $ rate_raw               : num  2.33 2.37 2.45 2.47 2.33 2.37 2.37 2.43 7.6 8.5 ...
 $ rate_bg                : num  0.167 0.15 0.15 0.183 0.183 0.233 0.167 0.183 0.233 0.233 ...
 $ rate_corrected         : num  2.17 2.22 2.3 2.28 2.15 2.13 2.2 2.25 7.37 8.27 ...
 $ activity_published     : num  37 66 56 35 21 ...
 $ cv_flag                : chr  "pass" "pass" "pass" "pass" ...
 $ interpretation         : chr  "usable" "usable" "usable" "usable" ...
 $ plate_date             : chr  "2026-08-14" "2026-08-14" "2026-08-14" "2026-08-14" ...
 $ std_slope_A412_per_nmol: num  0.0307 0.0307 0.0307 0.0307 0.0307 ...
 $ bg_undo_factor         : num  1.07 1.07 1.07 1.08 1.08 ...
 $ activity_nobg          : num  39.7 70.5 59.7 37.9 22.7 ...
 $ log10_activity         : num  1.6 1.85 1.78 1.58 1.36 ...

3.4 Independently verify the rescaling

The rescaling above is checked against a from-scratch recomputation of activity out of the raw rate and each plate’s own standard-curve slope, using the Abcam formula (nmol/min = (rate/1000) / std_slope; mU/uL = nmol/min / sample_volume * D; mU/mg = mU/uL * homogenate_volume / total_protein_mg). The two paths must agree to within the rounding of the input CSVs.

# Assay constants shared by all three plates (identical in all source analyses).
sample_volume_uL     <- 2
dilution_factor      <- 1
homogenate_volume_uL <- 350

verify <- cs_activity %>%
  mutate(
    activity_recomputed =
      ((rate_raw / 1000) / std_slope_A412_per_nmol / sample_volume_uL *
         dilution_factor) * homogenate_volume_uL / protein_mg,
    pct_diff = 100 * abs(activity_recomputed / activity_nobg - 1)
  )

cat(sprintf("Max %% difference, rescaled vs recomputed-from-raw: %.3f%%\n",
            max(verify$pct_diff)))
cat(sprintf("All within 1%% (input CSV rounding): %s\n", max(verify$pct_diff) < 1))
Max % difference, rescaled vs recomputed-from-raw: 0.415%
All within 1% (input CSV rounding): TRUE
cat(sprintf(
  "Removing the background correction changes activity by %.1f%%-%.1f%% (median %.1f%%).\n",
  100 * (min(cs_activity$bg_undo_factor) - 1),
  100 * (max(cs_activity$bg_undo_factor) - 1),
  100 * (median(cs_activity$bg_undo_factor) - 1)))
Removing the background correction changes activity by 0.0%-11.3% (median 0.8%).

4 ANALYSIS DATA SET

4.1 Set factor levels

# Group levels are ordered family-major so that, on a single x axis, each
# family's ambient and 36C boxes sit side by side.
group_levels <- c("F05 ambient", "F05 36C", "F07 ambient", "F07 36C")

cs_analysis <- cs_activity %>%
  mutate(
    family      = factor(family, levels = c("F05", "F07")),
    temperature = factor(temperature, levels = c("ambient", "36C")),
    group       = factor(paste(family, temperature), levels = group_levels),
    plate_dir   = factor(plate_dir)
  )

cat("--- cs_analysis: analysis-ready data with ordered factors ---\n")
cat("family levels:     ", paste(levels(cs_analysis$family), collapse = ", "), "\n")
cat("temperature levels:", paste(levels(cs_analysis$temperature), collapse = ", "), "\n")
cat("group levels:      ", paste(levels(cs_analysis$group), collapse = " | "), "\n")
cat("n rows:            ", nrow(cs_analysis), "\n")
--- cs_analysis: analysis-ready data with ordered factors ---
family levels:      F05, F07 
temperature levels: ambient, 36C 
group levels:       F05 ambient | F05 36C | F07 ambient | F07 36C 
n rows:             32

4.2 Colour palette

Colours are drawn from the Okabe-Ito qualitative palette, which is designed to remain distinguishable under deuteranopia, protanopia and tritanopia (Okabe & Ito 2008, Color Universal Design). One hue is bound to each family and reused in every figure in this document, so colour alone identifies the family throughout. Temperature is additionally encoded by point shape and by a light/dark tint of the family hue, so no figure relies on colour as its only channel – each remains readable in greyscale and to a colourblind reader.

# Okabe-Ito colourblind-safe qualitative palette.
okabe_ito <- c(
  blue          = "#0072B2",
  vermillion    = "#D55E00",
  bluishgreen   = "#009E73",
  orange        = "#E69F00",
  skyblue       = "#56B4E9",
  reddishpurple = "#CC79A7",
  yellow        = "#F0E442",
  black         = "#000000"
)

# Family identity: one hue per family.
family_colours <- c(F05 = okabe_ito[["blue"]], F07 = okabe_ito[["vermillion"]])

# Group identity: each family keeps its hue, with the 36C member of the pair
# as the lighter tint, so family and temperature are both readable from fill.
group_colours <- c(
  "F05 ambient" = okabe_ito[["blue"]],
  "F05 36C"     = okabe_ito[["skyblue"]],
  "F07 ambient" = okabe_ito[["vermillion"]],
  "F07 36C"     = okabe_ito[["orange"]]
)

stopifnot(all(levels(cs_analysis$group) %in% names(group_colours)))

cat("--- family_colours: hue bound to each family in every figure ---\n")
print(family_colours)
cat("\n--- group_colours: fill bound to each family x temperature group ---\n")
print(group_colours)
--- family_colours: hue bound to each family in every figure ---
      F05       F07 
"#0072B2" "#D55E00" 

--- group_colours: fill bound to each family x temperature group ---
F05 ambient     F05 36C F07 ambient     F07 36C 
  "#0072B2"   "#56B4E9"   "#D55E00"   "#E69F00"

4.3 QC status of the analysis set

Every individual is retained. The per-plate analyses already excluded unusable technical replicates before averaging; the flags below describe which individuals had a replicate dropped or a wide replicate CV, so any group difference can be checked against QC status rather than silently inheriting it.

qc_status <- cs_analysis %>%
  group_by(family, temperature) %>%
  summarise(
    n                   = n(),
    n_full_triplicate   = sum(reps_used == "3/3"),
    n_cv_flagged        = sum(cv_flag != "pass"),
    n_not_usable        = sum(interpretation != "usable"),
    max_cv_used_pct     = max(cv_used_pct),
    .groups = "drop"
  )

cat("--- qc_status: QC composition of each group ---\n")
print(as.data.frame(qc_status))
--- qc_status: QC composition of each group ---
  family temperature n n_full_triplicate n_cv_flagged n_not_usable
1    F05     ambient 8                 8            0            0
2    F05         36C 8                 8            0            0
3    F07     ambient 8                 7            1            0
4    F07         36C 8                 8            0            0
  max_cv_used_pct
1             4.6
2             7.8
3             6.4
4             3.1
qc_status %>%
  transmute(
    Family = family, Temperature = temperature, n = n,
    `Full 3/3 reps` = n_full_triplicate,
    `CV-flagged`    = n_cv_flagged,
    `Not usable`    = n_not_usable,
    `Worst CV of used reps (%)` = max_cv_used_pct
  ) %>%
  kable(caption = paste0("QC composition per group. CV-flagged counts",
                         " individuals whose CV across ALL replicates exceeded ",
                         analysis_params$cv_threshold_pct,
                         "%; such individuals still contribute a mean over",
                         " their retained replicates."))
Family Temperature n Full 3/3 reps CV-flagged Not usable Worst CV of used reps (%)
F05 ambient 8 8 0 0 4.6
F05 36C 8 8 0 0 7.8
F07 ambient 8 7 1 0 6.4
F07 36C 8 8 0 0 3.1

QC composition per group. CV-flagged counts individuals whose CV across ALL replicates exceeded 15%; such individuals still contribute a mean over their retained replicates.

5 DESCRIPTIVE STATISTICS

5.1 Per-group summary

group_summary <- cs_analysis %>%
  group_by(family, temperature) %>%
  summarise(
    n        = n(),
    mean_act = mean(activity_nobg),
    sd_act   = sd(activity_nobg),
    cv_act   = 100 * sd(activity_nobg) / mean(activity_nobg),
    median_act = median(activity_nobg),
    min_act  = min(activity_nobg),
    max_act  = max(activity_nobg),
    mean_rate_raw = mean(rate_raw),
    sd_rate_raw   = sd(rate_raw),
    .groups  = "drop"
  )

cat("--- group_summary: activity (mU/mg protein), background NOT subtracted ---\n")
print(as.data.frame(group_summary))
--- group_summary: activity (mU/mg protein), background NOT subtracted ---
  family temperature n  mean_act   sd_act   cv_act median_act  min_act
1    F05     ambient 8 114.20466 19.52111 17.09310  115.90009 84.22106
2    F05         36C 8  40.59051 16.27503 40.09566   36.28073 22.72780
3    F07     ambient 8 129.52200 40.91814 31.59165  124.21200 79.56600
4    F07         36C 8 129.02000 36.02833 27.92461  118.56250 86.54400
    max_act mean_rate_raw sd_rate_raw
1 139.34016      10.62875  3.23688315
2  70.49255       2.39000  0.05345225
3 207.81900       6.23000  1.62935570
4 200.57800       5.99750  2.05133928
group_summary %>%
  transmute(
    Family = family, Temperature = temperature, n = n,
    `Mean (mU/mg)`   = round(mean_act, 1),
    `SD`             = round(sd_act, 1),
    `CV (%)`         = round(cv_act, 1),
    `Median`         = round(median_act, 1),
    `Range`          = sprintf("%.1f - %.1f", min_act, max_act),
    `Raw rate (mA412/min)` = sprintf("%.2f +/- %.2f", mean_rate_raw, sd_rate_raw)
  ) %>%
  kable(caption = paste("Citrate synthase activity per family x temperature",
                        "group, with background subtraction removed (note 2)."))
Family Temperature n Mean (mU/mg) SD CV (%) Median Range Raw rate (mA412/min)
F05 ambient 8 114.2 19.5 17.1 115.9 84.2 - 139.3 10.63 +/- 3.24
F05 36C 8 40.6 16.3 40.1 36.3 22.7 - 70.5 2.39 +/- 0.05
F07 ambient 8 129.5 40.9 31.6 124.2 79.6 - 207.8 6.23 +/- 1.63
F07 36C 8 129.0 36.0 27.9 118.6 86.5 - 200.6 6.00 +/- 2.05

Citrate synthase activity per family x temperature group, with background subtraction removed (note 2).

5.2 Temperature effect within each family

temp_effect <- group_summary %>%
  select(family, temperature, mean_act) %>%
  pivot_wider(names_from = temperature, values_from = mean_act) %>%
  mutate(
    abs_change    = `36C` - ambient,
    fold_change   = `36C` / ambient,
    pct_change    = 100 * (fold_change - 1)
  )

cat("--- temp_effect: ambient -> 36C change in mean activity, per family ---\n")
print(as.data.frame(temp_effect))
--- temp_effect: ambient -> 36C change in mean activity, per family ---
  family  ambient       36C abs_change fold_change  pct_change
1    F05 114.2047  40.59051  -73.61414   0.3554191 -64.4580931
2    F07 129.5220 129.02000   -0.50200   0.9961242  -0.3875789
temp_effect %>%
  transmute(
    Family = family,
    `Ambient mean (mU/mg)` = round(ambient, 1),
    `36C mean (mU/mg)`     = round(`36C`, 1),
    `Change (mU/mg)`       = round(abs_change, 1),
    `Fold change`          = round(fold_change, 2),
    `% change`             = round(pct_change, 1)
  ) %>%
  kable(caption = "Direction and size of the 36C response within each family.")
Family Ambient mean (mU/mg) 36C mean (mU/mg) Change (mU/mg) Fold change % change
F05 114.2 40.6 -73.6 0.36 -64.5
F07 129.5 129.0 -0.5 1.00 -0.4

Direction and size of the 36C response within each family.

5.3 Individual-level table

individual_table <- cs_analysis %>%
  arrange(family, temperature, individual) %>%
  transmute(
    Sample = sample_id, Family = family, Individual = individual,
    Temperature = temperature,
    `Protein (mg)` = protein_mg,
    `Raw rate (mA412/min)` = rate_raw,
    `Activity, no BG (mU/mg)` = round(activity_nobg, 1),
    `Activity, published (mU/mg)` = activity_published,
    `Reps used` = reps_used,
    `CV used (%)` = cv_used_pct,
    `QC` = interpretation
  )

write.csv(individual_table,
          file.path(output_dir, "cs_activity_all_families_no_background.csv"),
          row.names = FALSE)

kable(individual_table,
      caption = paste("All 32 individuals with background subtraction",
                      "removed, alongside the originally published",
                      "background-corrected value for reference."))
Sample Family Individual Temperature Protein (mg) Raw rate (mA412/min) Activity, no BG (mU/mg) Activity, published (mU/mg) Reps used CV used (%) QC
F05_01_ambient F05 01 ambient 0.307 7.60 134.0 129.985 3/3 4.6 usable
F05_02_ambient F05 02 ambient 0.364 8.50 126.4 123.008 3/3 4.1 usable
F05_03_ambient F05 03 ambient 0.665 13.48 109.7 107.782 3/3 2.2 usable
F05_04_ambient F05 04 ambient 0.395 10.18 139.3 136.192 3/3 2.5 usable
F05_05_ambient F05 05 ambient 0.249 5.62 122.1 116.931 3/3 1.9 usable
F05_06_ambient F05 06 ambient 0.781 13.62 94.3 92.676 3/3 1.9 usable
F05_07_ambient F05 07 ambient 0.591 11.30 103.5 101.361 3/3 1.3 usable
F05_08_ambient F05 08 ambient 0.946 14.73 84.2 82.906 3/3 0.8 usable
F05_01_36C F05 01 36C 0.334 2.33 39.7 36.960 3/3 3.3 usable
F05_02_36C F05 02 36C 0.191 2.37 70.5 66.031 3/3 2.4 usable
F05_03_36C F05 03 36C 0.234 2.45 59.7 56.014 3/3 3.5 usable
F05_04_36C F05 04 36C 0.372 2.47 37.9 34.950 3/3 2.3 usable
F05_05_36C F05 05 36C 0.583 2.33 22.7 20.972 3/3 1.2 usable
F05_06_36C F05 06 36C 0.443 2.37 30.5 27.399 3/3 2.4 usable
F05_07_36C F05 07 36C 0.389 2.37 34.7 32.210 3/3 1.2 usable
F05_08_36C F05 08 36C 0.475 2.43 29.1 26.948 3/3 7.8 usable
F07_01_ambient F07 01 ambient 0.092 3.43 207.8 207.819 3/3 1.7 usable
F07_02_ambient F07 02 ambient 0.378 7.43 109.6 109.600 3/3 1.9 usable
F07_03_ambient F07 03 ambient 0.404 7.18 98.9 98.879 2/3 6.4 usable
F07_04_ambient F07 04 ambient 0.196 4.97 141.2 141.243 3/3 1.2 usable
F07_05_ambient F07 05 ambient 0.231 6.53 157.8 157.813 3/3 1.6 usable
F07_06_ambient F07 06 ambient 0.392 7.20 102.4 102.432 3/3 1.8 usable
F07_07_ambient F07 07 ambient 0.576 8.22 79.6 79.566 3/3 0.7 usable
F07_08_ambient F07 08 ambient 0.196 4.88 138.8 138.824 3/3 0.6 usable
F07_01_36C F07 01 36C 0.232 4.93 118.4 118.446 3/3 1.2 usable
F07_02_36C F07 02 36C 0.326 7.48 127.8 127.815 3/3 1.7 usable
F07_03_36C F07 03 36C 0.422 8.28 109.4 109.428 3/3 3.1 usable
F07_04_36C F07 04 36C 0.549 8.52 86.5 86.544 3/3 1.5 usable
F07_05_36C F07 05 36C 0.150 4.37 162.5 162.517 3/3 0.7 usable
F07_06_36C F07 06 36C 0.174 3.70 118.7 118.679 3/3 2.3 usable
F07_07_36C F07 07 36C 0.363 7.05 108.2 108.153 3/3 0.7 usable
F07_08_36C F07 08 36C 0.101 3.65 200.6 200.578 3/3 0.0 usable

All 32 individuals with background subtraction removed, alongside the originally published background-corrected value for reference.

6 STATISTICS

6.1 Two-way ANOVA on log10 activity

The model is log10(activity) ~ family * temperature. The interaction term is the formal test of the question in this document: does the temperature response differ between families? Because the design is balanced (8 per cell), sequential sums of squares are unambiguous.

cs_aov <- aov(log10_activity ~ family * temperature, data = cs_analysis)

cat("--- cs_aov: two-way factorial ANOVA on log10 activity ---\n")
print(summary(cs_aov))
--- cs_aov: two-way factorial ANOVA on log10 activity ---
                   Df Sum Sq Mean Sq F value   Pr(>F)    
family              1 0.6249  0.6249   39.89 7.85e-07 ***
temperature         1 0.4384  0.4384   27.99 1.25e-05 ***
family:temperature  1 0.4496  0.4496   28.70 1.04e-05 ***
Residuals          28 0.4386  0.0157                     
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
aov_tidy <- as.data.frame(summary(cs_aov)[[1]])
aov_tidy$Term <- trimws(rownames(aov_tidy))

anova_table <- aov_tidy %>%
  transmute(
    Term    = Term,
    Df      = Df,
    `Sum Sq`  = round(`Sum Sq`, 4),
    `Mean Sq` = round(`Mean Sq`, 4),
    `F value` = ifelse(is.na(`F value`), NA, round(`F value`, 3)),
    `p value` = ifelse(is.na(`Pr(>F)`), NA, signif(`Pr(>F)`, 4)),
    Significant = ifelse(is.na(`Pr(>F)`), "",
                         ifelse(`Pr(>F)` < analysis_params$alpha, "yes", "no"))
  )

kable(anova_table, row.names = FALSE,
      caption = sprintf(paste("Two-way ANOVA of log10 CS activity.",
                              "The family:temperature row is the test of",
                              "whether families differ in temperature",
                              "response (alpha = %.2f)."),
                        analysis_params$alpha))
Term Df Sum Sq Mean Sq F value p value Significant
family 1 0.6249 0.6249 39.892 8.00e-07 yes
temperature 1 0.4384 0.4384 27.988 1.25e-05 yes
family:temperature 1 0.4496 0.4496 28.702 1.04e-05 yes
Residuals 28 0.4386 0.0157 NA NA

Two-way ANOVA of log10 CS activity. The family:temperature row is the test of whether families differ in temperature response (alpha = 0.05).

p_family      <- aov_tidy$`Pr(>F)`[aov_tidy$Term == "family"]
p_temperature <- aov_tidy$`Pr(>F)`[aov_tidy$Term == "temperature"]
p_interaction <- aov_tidy$`Pr(>F)`[aov_tidy$Term == "family:temperature"]

cat(sprintf("family effect:              p = %.5g\n", p_family))
cat(sprintf("temperature effect:         p = %.5g\n", p_temperature))
cat(sprintf("family x temperature:       p = %.5g\n", p_interaction))
cat(sprintf("\nInteraction significant at alpha = %.2f: %s\n",
            analysis_params$alpha, p_interaction < analysis_params$alpha))
family effect:              p = 7.8529e-07
temperature effect:         p = 1.2544e-05
family x temperature:       p = 1.0448e-05

Interaction significant at alpha = 0.05: TRUE

6.2 Variance explained

ss <- aov_tidy$`Sum Sq`
names(ss) <- aov_tidy$Term
ss_total <- sum(ss)

eta_sq <- tibble(
  Term          = names(ss),
  `Sum Sq`      = round(ss, 4),
  `% of total variance` = round(100 * ss / ss_total, 1)
)

cat("--- eta_sq: partition of variance in log10 activity ---\n")
print(as.data.frame(eta_sq))
--- eta_sq: partition of variance in log10 activity ---
                Term Sum Sq % of total variance
1             family 0.6249                32.0
2        temperature 0.4384                22.5
3 family:temperature 0.4496                23.0
4          Residuals 0.4386                22.5
kable(eta_sq, caption = paste("Partition of total variance in log10 activity",
                              "(eta-squared). Residuals represent",
                              "between-individual variation within groups."))
Term Sum Sq % of total variance
family 0.6249 32.0
temperature 0.4384 22.5
family:temperature 0.4496 23.0
Residuals 0.4386 22.5

Partition of total variance in log10 activity (eta-squared). Residuals represent between-individual variation within groups.

6.3 Residual diagnostics

resid_shapiro <- shapiro.test(residuals(cs_aov))
resid_bartlett <- bartlett.test(log10_activity ~ group, data = cs_analysis)

cat(sprintf("Shapiro-Wilk on residuals:  W = %.4f, p = %.4g\n",
            resid_shapiro$statistic, resid_shapiro$p.value))
cat(sprintf("Bartlett homogeneity:       K^2 = %.4f, p = %.4g\n",
            resid_bartlett$statistic, resid_bartlett$p.value))
cat(sprintf("\nResiduals consistent with normality (p > %.2f): %s\n",
            analysis_params$alpha, resid_shapiro$p.value > analysis_params$alpha))
cat(sprintf("Group variances homogeneous (p > %.2f): %s\n",
            analysis_params$alpha, resid_bartlett$p.value > analysis_params$alpha))
Shapiro-Wilk on residuals:  W = 0.9765, p = 0.6941
Bartlett homogeneity:       K^2 = 3.5845, p = 0.31

Residuals consistent with normality (p > 0.05): TRUE
Group variances homogeneous (p > 0.05): TRUE
diag_df <- tibble(
  fitted   = fitted(cs_aov),
  residual = residuals(cs_aov),
  group    = cs_analysis$group
)

resid_fitted_plot <- ggplot(diag_df,
                            aes(x = fitted, y = residual,
                                colour = group, shape = group)) +
  geom_hline(yintercept = 0, linetype = "dashed", colour = "grey40") +
  geom_point(size = 2.5, alpha = 0.9) +
  scale_colour_manual(values = group_colours, name = "Group") +
  scale_shape_manual(values = c(16, 17, 15, 18), name = "Group") +
  labs(title = "Residuals vs fitted values",
       subtitle = "Two-way ANOVA on log10 CS activity",
       x = "Fitted log10 activity", y = "Residual") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"))

qq_df <- diag_df %>%
  arrange(residual) %>%
  mutate(theoretical = qnorm(ppoints(n())))

qq_plot <- ggplot(qq_df, aes(x = theoretical, y = residual)) +
  geom_abline(slope = sd(qq_df$residual), intercept = mean(qq_df$residual),
              linetype = "dashed", colour = "grey40") +
  geom_point(aes(colour = group, shape = group), size = 2.5, alpha = 0.9) +
  scale_colour_manual(values = group_colours, name = "Group") +
  scale_shape_manual(values = c(16, 17, 15, 18), name = "Group") +
  labs(title = "Normal Q-Q plot of residuals",
       subtitle = sprintf("Shapiro-Wilk p = %.3g", resid_shapiro$p.value),
       x = "Theoretical quantile", y = "Residual") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"))

ggsave(file.path(output_dir, "anova_residuals_vs_fitted.png"), resid_fitted_plot,
       width = 9, height = 4.5, dpi = 300)
ggsave(file.path(output_dir, "anova_qq_residuals.png"), qq_plot,
       width = 9, height = 4.5, dpi = 300)

print(resid_fitted_plot)

print(qq_plot)

6.4 Tukey HSD pairwise contrasts

tukey_res <- TukeyHSD(cs_aov, which = "family:temperature")

cat("--- tukey_res: all pairwise group contrasts on log10 activity ---\n")
print(tukey_res)
--- tukey_res: all pairwise group contrasts on log10 activity ---
  Tukey multiple comparisons of means
    95% family-wise confidence level

Fit: aov(formula = log10_activity ~ family * temperature, data = cs_analysis)

$`family:temperature`
                               diff        lwr        upr     p adj
F07:ambient-F05:ambient  0.04242003 -0.1284413  0.2132814 0.9046094
F05:36C-F05:ambient     -0.47116676 -0.6420281 -0.3003054 0.0000002
F07:36C-F05:ambient      0.04538564 -0.1254757  0.2162470 0.8861705
F05:36C-F07:ambient     -0.51358679 -0.6844482 -0.3427254 0.0000000
F07:36C-F07:ambient      0.00296561 -0.1678958  0.1738270 0.9999608
F07:36C-F05:36C          0.51655240  0.3456910  0.6874138 0.0000000
tukey_df <- as.data.frame(tukey_res$`family:temperature`)
tukey_df$Contrast <- rownames(tukey_df)

tukey_table <- tukey_df %>%
  transmute(
    Contrast = Contrast,
    `Diff (log10)`  = round(diff, 4),
    `Fold change`   = round(10^diff, 2),
    `95% CI (fold)` = sprintf("%.2f - %.2f", 10^lwr, 10^upr),
    `p adj`         = signif(`p adj`, 4),
    Significant     = ifelse(`p adj` < analysis_params$alpha, "yes", "no")
  ) %>%
  arrange(`p adj`)

write.csv(tukey_table, file.path(output_dir, "tukey_pairwise_contrasts.csv"),
          row.names = FALSE)

kable(tukey_table, row.names = FALSE,
      caption = paste("Tukey HSD contrasts between all four groups.",
                      "Differences on the log10 scale are back-transformed to",
                      "fold changes; a fold change of 1 means no difference."))
Contrast Diff (log10) Fold change 95% CI (fold) p adj Significant
F07:36C-F05:36C 0.5166 3.29 2.22 - 4.87 0.0000000 yes
F05:36C-F07:ambient -0.5136 0.31 0.21 - 0.45 0.0000000 yes
F05:36C-F05:ambient -0.4712 0.34 0.23 - 0.50 0.0000002 yes
F07:36C-F05:ambient 0.0454 1.11 0.75 - 1.65 0.8862000 no
F07:ambient-F05:ambient 0.0424 1.10 0.74 - 1.63 0.9046000 no
F07:36C-F07:ambient 0.0030 1.01 0.68 - 1.49 1.0000000 no

Tukey HSD contrasts between all four groups. Differences on the log10 scale are back-transformed to fold changes; a fold change of 1 means no difference.

6.5 The two within-family temperature contrasts

Pulled out of the Tukey table because these two rows, read together, are the interaction: how each family responds to 36 °C.

# TukeyHSD labels interaction levels with a ":" separator, e.g. "F05:36C".
within_family <- tukey_table %>%
  filter(Contrast %in% c("F05:36C-F05:ambient", "F07:36C-F07:ambient"))

stopifnot(nrow(within_family) == 2)

cat("--- within_family: the ambient vs 36C contrast inside each family ---\n")
print(as.data.frame(within_family))
--- within_family: the ambient vs 36C contrast inside each family ---
                               Contrast Diff (log10) Fold change 95% CI (fold)
F05:36C-F05:ambient F05:36C-F05:ambient      -0.4712        0.34   0.23 - 0.50
F07:36C-F07:ambient F07:36C-F07:ambient       0.0030        1.01   0.68 - 1.49
                        p adj Significant
F05:36C-F05:ambient 1.954e-07         yes
F07:36C-F07:ambient 1.000e+00          no
kable(within_family, row.names = FALSE,
      caption = "Within-family ambient vs 36C contrasts (Tukey-adjusted).")
Contrast Diff (log10) Fold change 95% CI (fold) p adj Significant
F05:36C-F05:ambient -0.4712 0.34 0.23 - 0.50 2e-07 yes
F07:36C-F07:ambient 0.0030 1.01 0.68 - 1.49 1e+00 no

Within-family ambient vs 36C contrasts (Tukey-adjusted).

7 FIGURES

7.1 Activity by family and temperature

Box plots of all four groups on a single axis, with every individual overlaid (n = 8 per group, so the underlying points are worth showing rather than summarising away). Boxes are the standard Tukey construction: the middle line is the median, the box spans the interquartile range, and the whiskers reach the most extreme point within 1.5 × IQR of the box.

7.1.1 Significance notation

The brackets are built from the Tukey HSD table computed above, not entered by hand, so the annotation cannot drift from the statistics.

Four of the six pairwise contrasts are annotated: the two within-family temperature contrasts and the two within-temperature family contrasts – i.e. every pair differing in exactly one factor. The two remaining “diagonal” contrasts (F05 ambient vs F07 36C, and F05 36C vs F07 ambient) change family and temperature at once and so cannot be attributed to either factor; they are not annotated. This is not selection by outcome – one omitted diagonal is significant and the other is not – and all six contrasts remain in the Tukey table and in tukey_pairwise_contrasts.csv. Non-significant retained contrasts are labelled ns rather than left blank.

# Star notation used on the plot.
sig_stars <- function(p) {
  ifelse(p < 0.001, "***",
  ifelse(p < 0.01,  "**",
  ifelse(p < 0.05,  "*", "ns")))
}

# x positions of each group on the plot's discrete axis.
group_x <- setNames(seq_along(group_levels), group_levels)

# TukeyHSD row names are "<groupB>-<groupA>" with ":" between factor levels.
sig_brackets <- tukey_df %>%
  transmute(
    contrast = Contrast,
    p_adj    = `p adj`,
    g2       = gsub(":", " ", sub("-.*$", "", Contrast)),
    g1       = gsub(":", " ", sub("^.*-", "", Contrast))
  ) %>%
  # Keep only pairs differing in exactly one factor (see prose above).
  mutate(
    fam1 = sub(" .*$", "", g1), fam2 = sub(" .*$", "", g2),
    tmp1 = sub("^\\S+ ", "", g1), tmp2 = sub("^\\S+ ", "", g2)
  ) %>%
  filter((fam1 == fam2) | (tmp1 == tmp2)) %>%
  mutate(
    xlo   = pmin(group_x[g1], group_x[g2]),
    xhi   = pmax(group_x[g1], group_x[g2]),
    span  = xhi - xlo,
    label = sig_stars(p_adj)
  ) %>%
  arrange(span, xlo)

stopifnot(nrow(sig_brackets) == 4)

# Pack brackets onto as few tiers as possible: shortest spans lowest, and two
# brackets share a tier when their x ranges do not overlap. This keeps the
# annotation band shallow so it does not crowd the data.
tier_of <- integer(nrow(sig_brackets))
tier_end <- numeric(0)  # right-most x reached on each tier so far
for (i in seq_len(nrow(sig_brackets))) {
  placed <- FALSE
  for (t in seq_along(tier_end)) {
    if (sig_brackets$xlo[i] > tier_end[t] + 0.25) {   # 0.25 = visual gap
      tier_of[i] <- t; tier_end[t] <- sig_brackets$xhi[i]; placed <- TRUE; break
    }
  }
  if (!placed) {
    tier_end <- c(tier_end, sig_brackets$xhi[i])
    tier_of[i] <- length(tier_end)
  }
}

y_data_max <- max(cs_analysis$activity_nobg)

sig_brackets <- sig_brackets %>%
  mutate(
    tier   = tier_of,
    y_top  = y_data_max * (1.05 + 0.075 * (tier - 1)),
    y_tick = y_data_max * 0.018
  )

cat("--- sig_brackets: significance annotations derived from tukey_df ---\n")
print(as.data.frame(sig_brackets[, c("contrast", "p_adj", "label",
                                     "xlo", "xhi", "tier")]))
cat(sprintf("\nBrackets packed onto %d tiers (from %d contrasts).\n",
            max(sig_brackets$tier), nrow(sig_brackets)))
--- sig_brackets: significance annotations derived from tukey_df ---
                                       contrast        p_adj label xlo xhi tier
F05:36C-F05:ambient         F05:36C-F05:ambient 1.953657e-07   ***   1   2    1
F07:36C-F07:ambient         F07:36C-F07:ambient 9.999608e-01    ns   3   4    1
F07:ambient-F05:ambient F07:ambient-F05:ambient 9.046094e-01    ns   1   3    2
F07:36C-F05:36C                 F07:36C-F05:36C 3.240667e-08   ***   2   4    3

Brackets packed onto 3 tiers (from 4 contrasts).
sig_caption <- paste0(
  "Significance: Tukey HSD on log10 activity. ",
  "*** p < 0.001,  ** p < 0.01,  * p < 0.05,  ns = not significant.\n",
  "Brackets show the four contrasts differing in exactly one factor; ",
  "all six pairwise contrasts are in tukey_pairwise_contrasts.csv.")

cat(sig_caption, "\n")
Significance: Tukey HSD on log10 activity. *** p < 0.001,  ** p < 0.01,  * p < 0.05,  ns = not significant.
Brackets show the four contrasts differing in exactly one factor; all six pairwise contrasts are in tukey_pairwise_contrasts.csv. 
set.seed(42)  # reproducible jitter

activity_plot <- ggplot(cs_analysis, aes(x = group, y = activity_nobg)) +
  geom_boxplot(aes(fill = group), outlier.shape = NA,
               width = 0.6, alpha = 0.55, linewidth = 0.5, colour = "grey25") +
  geom_jitter(aes(shape = temperature), width = 0.13, height = 0,
              size = 2.4, alpha = 0.9, colour = "grey15") +
  # Significance brackets: horizontal span + downward ticks + star label.
  geom_segment(data = sig_brackets,
               aes(x = xlo, xend = xhi, y = y_top, yend = y_top),
               inherit.aes = FALSE, linewidth = 0.4, colour = "grey30") +
  geom_segment(data = sig_brackets,
               aes(x = xlo, xend = xlo, y = y_top, yend = y_top - y_tick),
               inherit.aes = FALSE, linewidth = 0.4, colour = "grey30") +
  geom_segment(data = sig_brackets,
               aes(x = xhi, xend = xhi, y = y_top, yend = y_top - y_tick),
               inherit.aes = FALSE, linewidth = 0.4, colour = "grey30") +
  geom_text(data = sig_brackets,
            aes(x = (xlo + xhi) / 2, y = y_top, label = label),
            inherit.aes = FALSE, vjust = -0.3, size = 3.8) +
  scale_fill_manual(values = group_colours, guide = "none") +
  scale_shape_manual(values = c(ambient = 16, `36C` = 17),
                     name = "Temperature") +
  scale_y_continuous(expand = expansion(mult = c(0.04, 0.06))) +
  labs(
    title = "Family F05 loses citrate synthase activity at 36 C; family F07 does not",
    subtitle = sprintf(
      paste("Box = median and IQR, whiskers to 1.5 x IQR; each point is one",
            "individual (n = 8 per group). Background not subtracted.",
            "\nfamily x temperature p = %.2g"),
      p_interaction),
    x = NULL,
    y = "CS activity (mU / mg protein)",
    caption = sig_caption
  ) +
  theme_bw() +
  theme(plot.title    = element_text(size = 13, face = "bold"),
        plot.subtitle = element_text(size = 9.5),
        plot.caption  = element_text(size = 8.5, hjust = 0, colour = "grey30"),
        axis.text.x   = element_text(size = 11),
        legend.position = "right")

ggsave(file.path(output_dir, "cs_activity_by_family_temperature.png"), activity_plot,
       width = 10, height = 7, dpi = 300)

print(activity_plot)

7.2 Interaction plot

The two lines are the interaction: parallel lines would mean both families respond to temperature the same way. Family labels sit at the right end of each line, so no legend is needed.

interaction_df <- group_summary %>%
  mutate(
    family      = factor(family, levels = c("F05", "F07")),
    temperature = factor(temperature, levels = c("ambient", "36C")),
    se_act      = sd_act / sqrt(n),
    ci_half     = qt(0.975, df = n - 1) * se_act
  )

label_df <- interaction_df %>% filter(temperature == "36C")

interaction_plot <- ggplot(interaction_df,
                           aes(x = temperature, y = mean_act,
                               colour = family, group = family)) +
  geom_line(linewidth = 1.1) +
  geom_errorbar(aes(ymin = mean_act - ci_half, ymax = mean_act + ci_half),
                width = 0.07, linewidth = 0.7) +
  geom_point(size = 3.6) +
  geom_text(data = label_df, aes(label = family),
            hjust = -0.35, size = 4.2, fontface = "bold", show.legend = FALSE) +
  scale_colour_manual(values = family_colours, guide = "none") +
  scale_x_discrete(expand = expansion(mult = c(0.12, 0.28))) +
  labs(
    title = "The two families respond to 36 C differently (non-parallel lines)",
    subtitle = sprintf(
      "Group mean +/- 95%% CI, n = 8 per group. Background not subtracted. family x temperature p = %.2g",
      p_interaction),
    x = "Temperature exposure",
    y = "Mean CS activity (mU / mg protein)"
  ) +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"))

ggsave(file.path(output_dir, "cs_activity_interaction_plot.png"), interaction_plot,
       width = 9, height = 6, dpi = 300)

print(interaction_plot)

7.3 Per-individual activity

individual_plot <- ggplot(
  cs_analysis %>% mutate(label = paste0(family, "_", individual)),
  aes(x = reorder(label, activity_nobg), y = activity_nobg,
      colour = family, shape = temperature)) +
  geom_segment(aes(xend = reorder(label, activity_nobg), y = 0, yend = activity_nobg),
               linewidth = 0.4, colour = "grey75") +
  geom_point(size = 3) +
  scale_colour_manual(values = family_colours, name = "Family") +
  scale_shape_manual(values = c(ambient = 16, `36C` = 17), name = "Temperature") +
  coord_flip() +
  labs(
    title = "Per-individual CS activity across both families and exposures",
    subtitle = "Background not subtracted. Each row is one individual.",
    x = NULL,
    y = "CS activity (mU / mg protein)"
  ) +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.y = element_text(size = 7),
        legend.position = "right")

ggsave(file.path(output_dir, "cs_activity_per_individual.png"), individual_plot,
       width = 10, height = 6, dpi = 300)

print(individual_plot)

7.4 Effect of removing the background correction

A check that the decision in note 2 does not drive any group difference: if removing the correction mattered, points would sit far off the 1:1 line.

bg_plot <- ggplot(cs_analysis,
                  aes(x = activity_published, y = activity_nobg, colour = family)) +
  geom_abline(slope = 1, intercept = 0, linetype = "dashed", colour = "grey40") +
  geom_point(aes(shape = temperature), size = 2.8, alpha = 0.9) +
  scale_colour_manual(values = family_colours, name = "Family") +
  scale_shape_manual(values = c(ambient = 16, `36C` = 17), name = "Temperature") +
  labs(
    title = "Removing the background correction shifts activity by at most ~11%",
    subtitle = paste("Dashed line is 1:1 (no change). Points above the line are",
                     "the two F05 plates, where a background rate was estimable;",
                     "F07 points sit on the line."),
    x = "Activity as published, background subtracted (mU / mg protein)",
    y = "Activity with background NOT subtracted (mU / mg protein)"
  ) +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        plot.subtitle = element_text(size = 9))

ggsave(file.path(output_dir, "background_correction_effect.png"), bg_plot,
       width = 9, height = 6, dpi = 300)

print(bg_plot)

8 SUMMARY

f05_fold <- temp_effect$fold_change[temp_effect$family == "F05"]
f07_fold <- temp_effect$fold_change[temp_effect$family == "F07"]

f05_p <- within_family$`p adj`[within_family$Contrast == "F05:36C-F05:ambient"]
f07_p <- within_family$`p adj`[within_family$Contrast == "F07:36C-F07:ambient"]

amb_contrast <- tukey_table %>% filter(Contrast == "F07:ambient-F05:ambient")

cat(sprintf("F05: 36C / ambient = %.2f-fold (Tukey p = %.3g)\n", f05_fold, f05_p))
cat(sprintf("F07: 36C / ambient = %.2f-fold (Tukey p = %.3g)\n", f07_fold, f07_p))
cat(sprintf("Ambient F07 vs F05: %.2f-fold (Tukey p = %.3g)\n",
            amb_contrast$`Fold change`, amb_contrast$`p adj`))
cat(sprintf("Interaction p = %.3g | %% variance explained by interaction = %.1f%%\n",
            p_interaction,
            eta_sq$`% of total variance`[eta_sq$Term == "family:temperature"]))
F05: 36C / ambient = 0.36-fold (Tukey p = 1.95e-07)
F07: 36C / ambient = 1.00-fold (Tukey p = 1)
Ambient F07 vs F05: 1.10-fold (Tukey p = 0.905)
Interaction p = 1.04e-05 | % variance explained by interaction = 23.0%

8.1 Findings

cat(sprintf("
1. **The temperature response differs between the two families.** The
   family &times; temperature interaction is significant (p = %.2g), accounting
   for %.1f%% of the total variance in log10 activity. This is the central
   result: a single \"heat effect\" on CS activity does not describe both
   families.

2. **Family `F05` loses CS activity at 36 &deg;C.** Mean activity falls from
   %.1f to %.1f mU/mg protein, a **%.2f-fold change (%.0f%%)**, and the
   contrast survives Tukey adjustment (p = %.2g).

3. **Family `F07` shows essentially no temperature response.** Mean activity
   moves from %.1f to %.1f mU/mg protein -- a %.2f-fold change, entirely
   non-significant (p = %.2g). The `F07` 95%% CI at 36 &deg;C overlaps its
   ambient CI substantially.

4. **The two families are indistinguishable at ambient temperature**
   (%.2f-fold, p = %.2g). The families differ only *after* heat exposure,
   which is what makes this an interaction rather than a baseline family
   difference.

5. **Removing the background correction does not drive any of this.** Across
   all 32 individuals the correction was worth %.1f%%-%.1f%% of activity
   (median %.1f%%), and both families' ambient groups shift in the same
   direction, so no group contrast depends on the choice.
",
p_interaction,
eta_sq$`% of total variance`[eta_sq$Term == "family:temperature"],
temp_effect$ambient[temp_effect$family == "F05"],
temp_effect$`36C`[temp_effect$family == "F05"],
f05_fold, temp_effect$pct_change[temp_effect$family == "F05"], f05_p,
temp_effect$ambient[temp_effect$family == "F07"],
temp_effect$`36C`[temp_effect$family == "F07"],
f07_fold, f07_p,
amb_contrast$`Fold change`, amb_contrast$`p adj`,
100 * (min(cs_analysis$bg_undo_factor) - 1),
100 * (max(cs_analysis$bg_undo_factor) - 1),
100 * (median(cs_analysis$bg_undo_factor) - 1)))

  1. The temperature response differs between the two families. The family × temperature interaction is significant (p = 1e-05), accounting for 23.0% of the total variance in log10 activity. This is the central result: a single “heat effect” on CS activity does not describe both families.

  2. Family F05 loses CS activity at 36 °C. Mean activity falls from 114.2 to 40.6 mU/mg protein, a 0.36-fold change (-64%), and the contrast survives Tukey adjustment (p = 2e-07).

  3. Family F07 shows essentially no temperature response. Mean activity moves from 129.5 to 129.0 mU/mg protein – a 1.00-fold change, entirely non-significant (p = 1). The F07 95% CI at 36 °C overlaps its ambient CI substantially.

  4. The two families are indistinguishable at ambient temperature (1.10-fold, p = 0.9). The families differ only after heat exposure, which is what makes this an interaction rather than a baseline family difference.

  5. Removing the background correction does not drive any of this. Across all 32 individuals the correction was worth 0.0%-11.3% of activity (median 0.8%), and both families’ ambient groups shift in the same direction, so no group contrast depends on the choice.

8.2 Interpretation and caveats

cat(sprintf("
- **Statistical model.** Two-way ANOVA on log10 activity, balanced at n = 8
  per cell. Residuals are consistent with normality (Shapiro-Wilk
  p = %.3g) and group variances are homogeneous (Bartlett p = %.3g), so the
  parametric model is appropriate here. Working on the log scale makes the
  interaction a test of *fold-change* response, which is the meaningful
  comparison for an enzyme rate.

- **Assay design.** The `F05` ambient-vs-36 &deg;C comparison is made across
  two plates run ten days apart (2026-08-14 and 2026-08-24), whereas the
  `F07` comparison is made within a single plate (note 4). Cross-plate
  comparability is supported by the calibration and instrument-state values
  in [PLATE PROVENANCE](#plate-provenance): standard-curve slope varies by
  only %.1f%% CV across plates (all R<sup>2</sup> >= %.4f) and the CS
  positive control by %.1f%% CV.

- **Background correction.** No activity value in this document has
  background subtracted (note 2), so all four groups are treated
  identically. The `F07` plate's background wells received the wrong reagent
  and cannot be recovered; a future plate with correct Background Control Mix
  would let this choice be checked rather than assumed.

- **Scope.** Two families, eight individuals per family per exposure. These
  results describe these two families and do not establish how broadly
  family-dependent heat sensitivity extends across *M. gigas*.
",
resid_shapiro$p.value, resid_bartlett$p.value,
prov_spread$slope_cv, prov_spread$min_std_r2, prov_spread$pos_cv))

  • Statistical model. Two-way ANOVA on log10 activity, balanced at n = 8 per cell. Residuals are consistent with normality (Shapiro-Wilk p = 0.694) and group variances are homogeneous (Bartlett p = 0.31), so the parametric model is appropriate here. Working on the log scale makes the interaction a test of fold-change response, which is the meaningful comparison for an enzyme rate.

  • Assay design. The F05 ambient-vs-36 °C comparison is made across two plates run ten days apart (2026-08-14 and 2026-08-24), whereas the F07 comparison is made within a single plate (note 4). Cross-plate comparability is supported by the calibration and instrument-state values in PLATE PROVENANCE: standard-curve slope varies by only 2.5% CV across plates (all R2 >= 0.9983) and the CS positive control by 4.7% CV.

  • Background correction. No activity value in this document has background subtracted (note 2), so all four groups are treated identically. The F07 plate’s background wells received the wrong reagent and cannot be recovered; a future plate with correct Background Control Mix would let this choice be checked rather than assumed.

  • Scope. Two families, eight individuals per family per exposure. These results describe these two families and do not establish how broadly family-dependent heat sensitivity extends across M. gigas.

8.3 Output files

manifest <- tibble(
  file = list.files(output_dir),
) %>%
  mutate(
    size_kb = round(file.size(file.path(output_dir, file)) / 1024, 1),
    type    = case_when(
      grepl("\\.csv$", file) ~ "table",
      grepl("\\.png$", file) ~ "figure",
      TRUE                   ~ "documentation"
    )
  ) %>%
  arrange(type, file)

kable(manifest, caption = sprintf("Files written to `%s`.", output_dir))
file size_kb type
README.md 4.5 documentation
anova_qq_residuals.png 107.9 figure
anova_residuals_vs_fitted.png 95.8 figure
background_correction_effect.png 166.0 figure
cs_activity_by_family_temperature.png 157.4 figure
cs_activity_interaction_plot.png 132.1 figure
cs_activity_per_individual.png 134.6 figure
cs_activity_all_families_no_background.csv 2.6 table
tukey_pairwise_contrasts.csv 0.4 table

Files written to ../outputs/Gen5-20260825-mgig-sormi-citrate_synthase-F05-F07-temperature-comparison.

9 SESSION INFO

sessionInfo()
R version 4.6.1 (2026-06-24)
Platform: x86_64-pc-linux-gnu
Running under: Ubuntu 24.04.4 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/blas/libblas.so.3.12.0 
LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.12.0  LAPACK version 3.12.0

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

time zone: America/Los_Angeles
tzcode source: system (glibc)

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] tibble_3.3.1  ggplot2_4.0.3 tidyr_1.3.2   dplyr_1.2.1   knitr_1.51   

loaded via a namespace (and not attached):
 [1] vctrs_0.7.3        cli_3.6.6          rlang_1.2.0        xfun_0.57         
 [5] purrr_1.2.2        generics_0.1.4     textshaping_1.0.5  S7_0.2.2          
 [9] labeling_0.4.3     glue_1.8.1         htmltools_0.5.9    ragg_1.5.2        
[13] scales_1.4.0       rmarkdown_2.31     grid_4.6.1         evaluate_1.0.5    
[17] fastmap_1.2.0      yaml_2.3.12        lifecycle_1.0.5    compiler_4.6.1    
[21] RColorBrewer_1.1-3 pkgconfig_2.0.3    rstudioapi_0.18.0  systemfonts_1.3.2 
[25] farver_2.1.2       digest_0.6.39      R6_2.6.1           tidyselect_1.2.1  
[29] pillar_1.11.1      magrittr_2.0.5     withr_3.0.2        tools_4.6.1       
[33] gtable_0.3.6
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase ABCAM https://robertslab.github.io/sams-notebook/posts/2026/2026-08-25-Citrate-Synthase-Analysis---SORMI-June-2026-M.gigas-Family-5-vs-Family-7-Temperature-Response/ Tue, 25 Aug 2026 07:00:00 GMT Citrate Synthase Assay - SORMI June 2026 M.gigas Family 5 Ambient Ctenidia Re-assay Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-24-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-5-Ambient-Ctenidia-Re-assay/

INTRO

This is a re-assay of the eight Magallana gigas (Pacific oyster) Family 5 ctenidia homogenates held at ambient temperature, using the Citrate Synthase Assay Kit (ABCAM; cat: ab239712). The original assay of these same homogenates on 20260811 (Notebook entry) had QC failures serious enough that its activity values could not be used — a failed GSH standard curve, standards above the photometric linear range, background control wells behaving as active reactions, and one sample (F05_05_ambient) with no clean replicate. This plate is the re-run called for by that entry.

Two things changed relative to the 20260811 run: a shorter read window (20 min / 11 reads, vs. 40 min / 21 reads) and a pooled background control design in place of paired per-sample backgrounds. Samples had been previously homogenized on 20260811 (Notebook entry), and protein concentrations for normalization were determined on 20260813 (Notebook entry) — the same homogenates and the same protein measurements used for the original assay.

Note

This re-assay succeeded and supersedes the 20260811 ambient results. No sample well, positive-control well, or background well is flagged for anything on this plate: no decreasing kinetics, no read glitches, no elevated starting absorbance, and all three pooled-background replicates are flat. All eight samples pass the 15% technical CV threshold (0.8-4.6%), the GSH standard curve has zero flagged outlier wells (R2 = 0.9983), and no replicate is excluded from any sample. This plate does not need re-assay.

One caution worth carrying forward: the positive control’s three replicates rise linearly (R2 0.997-1.000) but disagree in rate by 22.8% CV (D7 at 2.00 mA412/min vs. D8/D9 at 2.95-3.15 mA412/min) — worse precision than any individual sample’s technical triplicate on this plate. This does not enter any activity calculation, so it is a plate-level caution rather than a disqualifier, but it is worth watching on future plates.

All data/code/analysis is available here:

MATERIALS & METHODS

Citrate Synthase Assay

The Citrate Synthase Assay Kit (ABCAM; cat: ab239712) was used according to the manufacturer’s instructions (ABCAM-Citrate-Synthase-Assay-v4a-ab239712.pdf (GitHub; PDF)). The kit is a coupled kinetic assay: citrate synthase condenses acetyl-CoA and oxaloacetate to release free CoA-SH, the liberated thiol reduces DTNB to TNB2-, and the rate of absorbance increase at 412 nm is proportional to citrate synthase activity. Absolute nmol of thiol are assigned from a reduced glutathione (GSH) standard curve read on the same plate.

Per reaction well, 2 µL of undiluted homogenate was combined with 48 µL of Assay Buffer 7, followed by 50 µL of Reaction Mix. Each of the eight samples was assayed in technical triplicate in a clear, flat-bottomed 96-well plate. A six-point GSH standard curve (0, 8, 16, 24, 32, 40 nmol/well) and a CS positive control were included in triplicate on the same plate.

REACTION MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 43 55 2365
DTNB 5 55 275
CS Substrate 2 55 110

BACKGROUND MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 45 4 180
DTNB 5 4 20

Absorbance was read at 412 nm in kinetic mode at 25°C on a Synergy HTX plate reader (Agilent), every 2 minutes for 20 minutes (11 reads).

Normalization

Activity was normalized to total extracted protein rather than tissue weight. Tissue was homogenized in 350 µL of Assay Buffer 7, and protein concentration in each homogenate was quantified on 20260813 (Notebook entry) using the Bradford assay. Total protein mass per sample is that concentration multiplied by the 350 µL homogenization volume. Tissue weights are recorded in the plate layout labels for provenance only and play no role in the activity calculation. Protein values come from two assays run on 20260813: the F05 plate (F05_04, F05_05, F05_07) and the F05/F07 re-assay plate (F05_01, F05_02, F05_03, F05_06, F05_08, re-assayed at dilution).

Analysis

Data were analyzed in R. The Rmd file is here (GitHub):

Background correction used a single pooled background control rather than a paired per-sample background as on the 20260811 and 20260814 plates. A pool was made from 5 µL of each of the eight sample homogenates, and 2 µL of that pool (matching the 2 µL sample volume used for every reaction well) was run in triplicate with Background Control Mix (no CS Substrate Mix), producing wells BG-citrate_synthase (D10-D12). Because there is only one background estimate for the whole plate, the same background-corrected rate is subtracted from every sample.

Knitted markdown is below.

Data files

Plate reader files

Plate reader files (*.xpt) are available here (require Gen5 software to open):

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/plate_reader_files

Raw absorbance data

Raw absorbance data (*.csv), plate layouts, and Gen5 full reports are available here:

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/raw_absorbance

RESULTS

Unlike the Family 5 36°C plate (Notebook entry), where activity was essentially uniform across individuals, this ambient plate shows real, well-resolved inter-individual variation. Background-corrected rates range 5.38 to 14.50 mA412/min across the eight individuals — a 169% spread against 0.8-4.6% technical CV per sample, far larger than the technical noise floor.

Protein normalization looks appropriate here rather than artifactual: raw rate (mU/µL) is strongly and significantly correlated with total extracted protein (r = +0.96, p = 0.0002), the opposite of the pattern on the 20260814 36°C plate. That makes mU/mg protein the recommended basis for downstream comparison on this plate — normalizing is removing a real extraction-efficiency covariate, not introducing an artifact.

Sample Reps used CV all reps (%) Corrected rate (mA412/min) Activity (mU/uL) Activity (mU/mg protein) Interpretation
F05_01_ambient 3/3 4.6 7.37 0.1139 129.985 usable
F05_02_ambient 3/3 4.1 8.27 0.1278 123.008 usable
F05_03_ambient 3/3 2.2 13.25 0.2048 107.782 usable
F05_04_ambient 3/3 2.5 9.95 0.1538 136.192 usable
F05_05_ambient 3/3 1.9 5.38 0.0832 116.931 usable
F05_06_ambient 3/3 1.9 13.38 0.2069 92.676 usable
F05_07_ambient 3/3 1.3 11.07 0.1711 101.361 usable
F05_08_ambient 3/3 0.8 14.50 0.2242 82.906 usable

Note that F05_05_ambient — the sample disqualified entirely on 20260811, where all three replicates started at an elevated A412 (0.65-0.93) indicating contamination or mis-dispensing rather than enzymology — now has three clean replicates at 1.9% CV and no elevated baseline. It is the lowest of the eight at 5.38 mA412/min.

The pooled background design cut the wells needed for background control roughly 8-fold (3 wells vs. 24 on the paired per-sample design). On this plate the background rate is a small fraction of the sample signal (~2.2%) either way, so the simplification looks low-risk here, though a pooled design should be re-validated if background ever becomes a larger fraction of the raw signal.

There is also widespread but harmless GSH standard signal decay: 13 of 15 (strict) to 15 of 18 (including exact-zero ties) non-zero-concentration standard wells lose signal between t = 0 and t = 20 min, consistent with TNB2- instability. Because the standard curve is calibrated at t = 0, this has no effect on the fit.

AI Use Level 4: Substantial AI contribution

1 BACKGROUND

Citrate synthase (CS) activity in ctenidia of eight Magallana gigas (Pacific oyster) individuals from family F05 held at ambient temperature, re-assayed 2026-08-24 with the Abcam Citrate Synthase Assay Kit (ab239712), v4a. This is the same eight F05 ambient homogenates originally assayed 2026-08-11 (Gen5-20260811-mgig-citrate_synthase-F05-ambient.Rmd), re-run with a shorter read window and a different background-control design (see note 3).

The kit is a coupled kinetic assay. CS condenses acetyl-CoA and oxaloacetate, releasing free CoA-SH; the liberated thiol reduces DTNB to TNB2-, which absorbs at 412 nm. The rate of A412 increase is proportional to CS activity, and absolute nmol of thiol are assigned from a GSH (reduced glutathione) standard curve read on the same plate.

Absorbance was read at 412 nm in kinetic mode, 25 °C, every 2 min for 20 min (11 reads) on a Synergy HTX. See note 1.

Total extracted protein per homogenate (used for normalization, note 2) was not measured on this plate; it comes from two BCA/Bradford-style protein-quantification assays run 2026-08-13: Gen5-20260813-mgig-sormi-BSA-F05-protein.Rmd (F05_04_ambient, F05_05_ambient, F05_07_ambient) and Gen5-20260813-mgig-sormi-BSA-F05-F07-reassay-protein.Rmd (F05_01_ambient, F05_02_ambient, F05_03_ambient, F05_06_ambient, F05_08_ambient, re-assayed at dilution). These are the same homogenates and the same protein measurements used for the original 2026-08-11 assay of this plate.

This document is fully self-contained: layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are all defined directly below (see ## Pipeline functions under SETUP) rather than sourced from a separate script, so the whole analysis can be reproduced from this one file.

1.1 Sample naming convention

Well labels follow <family>_<individual>_<temperature>-<assay_type>-<weight>-df.<n> for sample wells (e.g. F05_01_ambient-citrate_synthase-10.1-df.0), STD-<assay_type>-<nmol_per_well> for GSH standards, POS-<assay_type> for the positive control, and BG-<assay_type> for the pooled background control (see note 3) – the background label carries no weight or dilution suffix, since it is not tied to any single individual’s tissue. The trailing number in the sample label (e.g. 10.1) is tissue weight in mg, which is not used for normalization here (note 2); it is parsed and validated but plays no role in the activity calculation.

1.2 Notes

  1. Read duration for this plate is 20 minutes (11 reads at 2-min intervals) – shorter than the original 2026-08-11 run of these same homogenates (40 min, 21 reads). The assay parameters below set read_duration_min = 20 to match. Rate extraction uses a fixed-width sliding window (rate_window_n = 5 points = 8 min) rather than the full trace, so this only changes how many windows are available to slide across; see the KINETIC TRACES section for whether the assay is still in its linear phase over this window.
  2. Activity is normalized to total extracted protein, not tissue weight, because protein concentration was measured directly (BCA-style assay, sample_protein_concentrations_*.csv), and controls for extraction efficiency in a way tissue weight cannot. These are the same protein measurements used for the original 2026-08-11 ambient plate (same homogenates), and no Sample string in either protein-concentration CSV needed correction for this plate.
  3. Background correction uses a single pooled background control, not a paired per-sample background as on the 2026-08-11 and 2026-08-14 plates. A pool was made from 5 µL of each of the 8 sample homogenates, and 2 µL of that pool (matching sample_volume_uL used for every reaction well) was run in triplicate with Background Control Mix (no CS Substrate Mix) to produce wells BG-citrate_synthase (D10-D12). Because there is only one background estimate for the whole plate, the same background-corrected rate is subtracted from every sample, rather than each sample carrying its own paired background replicate set.
library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

1.3 Pipeline functions

All layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are defined directly in this document (rather than sourced from a separate script) so that it is fully self-contained and reproducible from this single file. The functions are grouped below by pipeline stage; each is called in the DATA / KINETIC TRACES / ANOMALY DETECTION / GSH STANDARD CURVE / RATE EXTRACTION / BACKGROUND CORRECTION / TECHNICAL REPLICATE PRECISION / CITRATE SYNTHASE ACTIVITY / QC SUMMARY sections that follow. Nothing plate-specific (typo fixes, disqualified samples, thresholds) is hardcoded in any function – it is always passed in as an argument, from the assay_params list defined further below or from a plate-specific fix list.

1.3.1 Data import and layout parsing

Reads the raw Gen5 exports (kinetic absorbance CSV, full-report text backup, plate-layout CSV), reconciles them against each other, and attaches well-type/sample-id/replicate metadata plus measured protein concentrations to every well.

## -----------------------------------------------------------------------
## elapsed_to_min()
## Purpose : Convert Gen5 "H:MM:SS" elapsed-time strings to decimal minutes.
## Inputs  : x - character vector of "H:MM:SS" strings
## Outputs : numeric vector of elapsed minutes, same length as x
## -----------------------------------------------------------------------
elapsed_to_min <- function(x) {
  vapply(strsplit(x, ":"), function(p) {
    p <- as.numeric(p)
    p[1] * 60 + p[2] + p[3] / 60
  }, numeric(1))
}

## -----------------------------------------------------------------------
## read_absorbance_csv()
## Purpose : Parse a Gen5 "absorbance-*.csv" kinetic export: wavelength
##           header, a "Time," row, a temperature row, then one row per well
##           (well ID in column 1, one OD reading per timepoint thereafter).
## Inputs  : path - path to the absorbance-*.csv file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_absorbance_csv <- function(path) {
  ln <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  time_i   <- grep("^Time,", ln)[1]
  time_str <- strsplit(ln[time_i], ",")[[1]]
  time_str <- time_str[!time_str %in% c("", "Time")]
  tmin     <- elapsed_to_min(time_str)

  well_ln <- ln[grepl("^[A-H][0-9]{1,2},", ln)]
  bind_rows(lapply(strsplit(well_ln, ","), function(p) {
    data.frame(well = p[1], time_min = tmin,
               od = as.numeric(p[seq_along(tmin) + 1]),
               stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## read_full_report()
## Purpose : Parse a Gen5 "full_report-*.txt" export: tab-delimited, wells
##           across columns, timepoints down rows. Used only as an
##           independent cross-check against the absorbance CSV, never as a
##           primary data source.
## Inputs  : path - path to the full_report-*.txt file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_full_report <- function(path) {
  ln  <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  hdr_i <- grep("^Time\t.*\tA1\t", ln)[1]
  hdr   <- trimws(strsplit(ln[hdr_i], "\t")[[1]])
  rows  <- ln[grepl("^[0-9]:[0-9]{2}:[0-9]{2}\t", ln)]
  m     <- do.call(rbind, strsplit(rows, "\t"))
  tmin  <- elapsed_to_min(m[, 1])

  well_cols <- which(grepl("^[A-H][0-9]{1,2}$", hdr))
  bind_rows(lapply(well_cols, function(j) {
    data.frame(well = hdr[j], time_min = tmin,
               od = as.numeric(m[, j]), stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## parse_plate_layout()
## Purpose : Convert a raw Gen5 plate-layout CSV export into one row per
##           occupied well with its descriptive label. Auto-detects two
##           export formats seen in practice:
##             "single"  - one row per plate row (row 1 = column numbers,
##                         column 1 = row letter, last column = "Name" tag);
##                         each cell already holds the full descriptive label.
##             "double"  - each plate row is exported as TWO rows: a
##                         "Well ID" row (SPL codes) immediately followed by
##                         a "Name" row (the descriptive label actually used
##                         downstream).
##           The format is detected from column 1: if a value repeats twice
##           in a row (once for the well-ID row, once for the label row)
##           immediately below a numeric header, it is treated as "double".
## Inputs  : path - path to the layout-*.csv file
## Outputs : data.frame, one row per occupied well:
##           well (chr, e.g. "A1"), plate_row (chr), plate_col (int),
##           label (chr, trimmed descriptive label)
## -----------------------------------------------------------------------
parse_plate_layout <- function(path) {
  plate_layout <- read.csv(path, header = FALSE, stringsAsFactors = FALSE)

  col1 <- trimws(plate_layout$V1)
  # "double" format: row letters appear twice in a row in column 1
  # (once tagging the Well-ID row, once tagging the Name row), e.g.
  # "A", "A", "B", "B", ... Detect by checking whether non-empty,
  # non-header values in col1 are each immediately followed by a repeat.
  data_rows   <- which(col1 != "" & !is.na(col1))[-1]  # drop header row (row 1)
  is_double <- length(data_rows) >= 2 &&
    all(col1[data_rows[c(TRUE, FALSE)]] == col1[data_rows[c(FALSE, TRUE)]][seq_len(floor(length(data_rows) / 2))])

  if (is_double) {
    # Pair each row-letter row with the following label row.
    row_letter_rows <- data_rows[c(TRUE, FALSE)]
    label_rows      <- data_rows[c(FALSE, TRUE)]
    stopifnot(length(row_letter_rows) == length(label_rows))

    out <- expand.grid(pair_idx = seq_along(row_letter_rows),
                        col_idx  = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = col1[row_letter_rows[pair_idx]],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j],
                                   label_rows[pair_idx], col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  } else {
    # "single" format: row 1 = column numbers, column 1 = row letters,
    # last column = "Name" tag; each cell already holds the full label.
    out <- expand.grid(row_idx = 2:nrow(plate_layout),
                        col_idx = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = plate_layout$V1[row_idx],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j], row_idx, col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  }

  attr(out, "layout_format") <- ifelse(is_double, "double", "single")
  out
}

## -----------------------------------------------------------------------
## reconcile_kinetic_sources()
## Purpose : Cross-check the absorbance CSV (primary data source) against
##           the full_report text export (independent instrument export)
##           and the plate layout (expected well set). Fails loudly
##           (stopifnot) if any layout well is missing from the CSV, or if
##           the two kinetic sources disagree on any shared reading.
## Inputs  : absorbance_csv - long-format data.frame from read_absorbance_csv()
##           full_report    - long-format data.frame from read_full_report()
##           layout_wells   - data.frame from parse_plate_layout()
##           tolerance      - max allowed |CSV - report| absolute difference
##                            (default 1e-9, i.e. exact agreement)
## Outputs : list(plate_readings, overlap_check, summary), where
##           plate_readings = absorbance_csv with a `source` column added,
##           summary        = named list of counts for reporting
## -----------------------------------------------------------------------
reconcile_kinetic_sources <- function(absorbance_csv, full_report, layout_wells,
                                       tolerance = 1e-9) {
  wells_csv    <- unique(absorbance_csv$well)
  wells_report <- unique(full_report$well)

  wells_report_not_csv <- setdiff(wells_report, wells_csv)
  wells_layout_not_csv <- setdiff(layout_wells$well, wells_csv)
  wells_csv_not_layout <- setdiff(wells_csv, layout_wells$well)

  overlap_check <- inner_join(absorbance_csv, full_report,
                              by = c("well", "time_min"), suffix = c("_csv", "_report")) %>%
    mutate(abs_diff = abs(od_csv - od_report))

  max_disagreement <- if (nrow(overlap_check) > 0) max(overlap_check$abs_diff) else NA_real_

  stopifnot(
    length(wells_layout_not_csv) == 0,
    is.na(max_disagreement) || max_disagreement < tolerance
  )

  plate_readings <- absorbance_csv %>% mutate(source = "absorbance_csv")

  list(
    plate_readings = plate_readings,
    overlap_check  = overlap_check,
    summary = list(
      n_wells_csv               = length(wells_csv),
      n_wells_report            = length(wells_report),
      n_wells_report_not_csv    = length(wells_report_not_csv),
      n_wells_layout_not_csv    = length(wells_layout_not_csv),
      n_wells_csv_not_layout    = length(wells_csv_not_layout),
      n_shared_readings         = nrow(overlap_check),
      max_disagreement          = max_disagreement
    )
  )
}

## -----------------------------------------------------------------------
## annotate_wells()
## Purpose : Join layout labels onto the kinetic readings and parse each
##           label into well_type, sample_id, family, individual,
##           temperature, tissue weight, and standard concentration.
##           Expected label grammar:
##             sample              : <family>_<individual>_<temperature>-<assay_type>-<weight>-df.<n>
##             standard            : STD-<assay_type>-<nmol_per_well>
##             positive control    : POS-<assay_type>
##             background (pooled) : BG-<assay_type>
##           Unlike the paired per-sample background design used on earlier
##           plates (`_BG` appended to a sample's own label), this plate's
##           background wells are a single pooled control with no sample_id,
##           family, individual, or weight of their own (note 3) -- so
##           sample_id/family/individual/temperature/weight_mg parsing is
##           restricted to well_type == "sample" only.
##           Label typos (e.g. a misspelled assay_type) are corrected before
##           parsing via `label_fixes`, a named character vector of
##           c(pattern = replacement) pairs applied in order with gsub().
## Inputs  : plate_readings - data.frame from reconcile_kinetic_sources()$plate_readings
##           layout_wells   - data.frame from parse_plate_layout()
##           label_fixes    - optional named character vector, e.g.
##                            c("citrate_synthasse" = "citrate_synthase")
## Outputs : plate_long - fully annotated long-format data.frame, one row
##           per well x timepoint, with well_type, sample_id, family,
##           individual, temperature, weight_mg, std_nmol columns added
## -----------------------------------------------------------------------
annotate_wells <- function(plate_readings, layout_wells, label_fixes = NULL) {
  plate_long <- plate_readings %>%
    left_join(layout_wells, by = "well") %>%
    mutate(
      well_type = case_when(
        grepl("^STD-", label) ~ "standard",
        grepl("^POS-", label) ~ "positive_control",
        grepl("^BG-",  label) ~ "background",
        TRUE                  ~ "sample"
      ),
      label_clean = label
    )

  if (!is.null(label_fixes)) {
    for (pat in names(label_fixes)) {
      plate_long$label_clean <- gsub(pat, label_fixes[[pat]], plate_long$label_clean)
    }
  }

  plate_long <- plate_long %>%
    mutate(
      sample_id   = ifelse(well_type == "sample",
                           sub("-[a-zA-Z_]+-.*$", "", label_clean), NA_character_),
      family      = ifelse(!is.na(sample_id), sub("^(F[0-9]+)_.*$", "\\1", sample_id), NA_character_),
      individual  = ifelse(!is.na(sample_id), sub("^F[0-9]+_([0-9]+)_.*$", "\\1", sample_id), NA_character_),
      temperature = ifelse(!is.na(sample_id), sub("^F[0-9]+_[0-9]+_(.*)$", "\\1", sample_id), NA_character_),
      weight_mg   = ifelse(well_type == "sample",
                           as.numeric(sub("^.*-[a-zA-Z_]+-([0-9.]+)-df\\..*$", "\\1", label_clean)),
                           NA_real_),
      std_nmol    = ifelse(well_type == "standard",
                           as.numeric(sub("^STD-[a-zA-Z_]+-", "", label_clean)), NA_real_)
    ) %>%
    arrange(well_type, sample_id, well, time_min)

  # Fail loudly rather than silently dropping malformed labels
  stopifnot(
    !any(is.na(plate_long$well_type)),
    !any(is.na(plate_long$weight_mg[plate_long$well_type == "sample"])),
    !any(is.na(plate_long$std_nmol[plate_long$well_type == "standard"]))
  )

  plate_long
}

## -----------------------------------------------------------------------
## load_protein_concentrations()
## Purpose : Load one or more BCA/Bradford-style protein-concentration CSV
##           exports, concatenate them, and match each plate sample to
##           exactly one protein-concentration record. Fails loudly if any
##           plate sample has zero or more than one match. Converts
##           concentration (ug/mL) to total protein mass (mg) in the full
##           homogenate using the measured homogenization volume.
## Inputs  : protein_files        - character vector of CSV paths, each with
##                                   columns `Sample` and
##                                   `Calculated concentration (ug/mL)`
##           plate_sample_ids     - character vector of this plate's sample IDs
##           homogenate_volume_uL - measured homogenization volume (uL),
##                                   applied identically to every sample
##           sample_id_fixes      - optional named character vector of
##                                   c(pattern = replacement) label-typo
##                                   corrections applied to the protein
##                                   file's Sample column before matching
## Outputs : data.frame, one row per plate sample:
##           sample_id, source_file, conc_ug_mL, total_protein_mg
## -----------------------------------------------------------------------
load_protein_concentrations <- function(protein_files, plate_sample_ids,
                                         homogenate_volume_uL,
                                         sample_id_fixes = NULL) {
  protein_all <- protein_files %>%
    purrr::map_dfr(~ read.csv(.x, check.names = FALSE) %>%
                      mutate(source_file = basename(.x))) %>%
    rename(sample_id_protein = Sample, conc_ug_mL = `Calculated concentration (ug/mL)`) %>%
    select(sample_id_protein, source_file, conc_ug_mL)

  if (!is.null(sample_id_fixes)) {
    for (pat in names(sample_id_fixes)) {
      protein_all$sample_id_protein <- gsub(pat, sample_id_fixes[[pat]],
                                             protein_all$sample_id_protein)
    }
  }

  protein_matches <- protein_all %>% filter(sample_id_protein %in% plate_sample_ids)
  match_counts    <- protein_matches %>% count(sample_id_protein, name = "n_matches")

  stopifnot(
    all(plate_sample_ids %in% protein_matches$sample_id_protein),
    all(match_counts$n_matches == 1)
  )

  protein_matches %>%
    rename(sample_id = sample_id_protein) %>%
    mutate(total_protein_mg = conc_ug_mL * homogenate_volume_uL / 1e6) %>%
    arrange(sample_id)
}

1.3.2 Well-level kinetic QC

Flags read glitches, non-rising (‘decreasing’) traces, elevated t0 baselines, and other well-level anomalies before any rate is trusted.

## -----------------------------------------------------------------------
## local_step_excess()
## Purpose : For a kinetic trace, measure how much each read-to-read step
##           departs from the AVERAGE of its two immediate neighbouring
##           steps. A large local excess flags a read glitch; comparing to
##           local neighbours (rather than the trace's overall median step)
##           tolerates traces that smoothly decelerate throughout.
## Inputs  : od - numeric vector of OD readings in time order
## Outputs : numeric vector of length length(od) - 1 (one per step)
## -----------------------------------------------------------------------
local_step_excess <- function(od) {
  d <- diff(od)
  n <- length(d)
  local_expect <- vapply(seq_len(n), function(i) {
    mean(d[intersect(c(i - 1L, i + 1L), seq_len(n))])
  }, numeric(1))
  d - local_expect
}

## -----------------------------------------------------------------------
## compute_well_diagnostics()
## Purpose : Compute four independent per-well trace-shape diagnostics from
##           the raw kinetic trace, before any rate fitting: direction
##           (net change), monotonicity (fraction of rising steps),
##           over-range (any read above the photometric linearity ceiling),
##           and discontinuity (a read-to-read glitch via local_step_excess).
##           Also flags an elevated starting absorbance for sample wells,
##           and background wells that behave like active reactions.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with od_linear_max, glitch_excess_od,
##                          sample_baseline_od, bg_flat_od_max,
##                          bg_flat_drift_max
## Outputs : well_diagnostics - one row per well, with od_first, od_last,
##           od_max, net_change, frac_rising, max_step, typical_step,
##           step_excess, glitch_at_min, and flag_* / n_flags columns
## -----------------------------------------------------------------------
compute_well_diagnostics <- function(plate_long, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, plate_row, plate_col, well_type, sample_id, label, std_nmol, source) %>%
    summarise(
      od_first        = first(od),
      od_last         = last(od),
      od_max          = max(od),
      net_change      = last(od) - first(od),
      frac_rising     = mean(diff(od) > 0),
      max_step        = diff(od)[which.max(abs(diff(od)))],
      typical_step    = median(abs(diff(od))),
      step_excess     = local_step_excess(od)[which.max(abs(local_step_excess(od)))],
      glitch_at_min   = time_min[which.max(abs(local_step_excess(od))) + 1L],
      .groups = "drop"
    ) %>%
    mutate(
      step_ratio      = abs(max_step) / pmax(typical_step, 0.001),
      flag_decreasing = net_change <= 0,
      flag_over_range = od_max > assay_params$od_linear_max,
      flag_glitch     = abs(step_excess) > assay_params$glitch_excess_od,
      flag_high_baseline = well_type == "sample" & od_first > assay_params$sample_baseline_od,
      flag_bg_active  = well_type == "background" &
                        (od_first > assay_params$bg_flat_od_max |
                         abs(net_change) > assay_params$bg_flat_drift_max),
      n_flags         = flag_decreasing + flag_over_range + flag_glitch +
                        flag_bg_active + flag_high_baseline
    )
}

## -----------------------------------------------------------------------
## compute_baseline_diagnostics()
## Purpose : Classify sample/background wells by starting absorbance
##           (elevated vs. normal) and summarize, per sample, how many of
##           its replicates are baseline-compromised. A sample with all
##           replicates elevated has no clean replicate at all.
## Inputs  : well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with sample_baseline_od
## Outputs : list(baseline_check, baseline_per_sample)
## -----------------------------------------------------------------------
compute_baseline_diagnostics <- function(well_diagnostics, assay_params) {
  baseline_check <- well_diagnostics %>%
    filter(well_type %in% c("sample", "background")) %>%
    mutate(baseline = ifelse(od_first > assay_params$sample_baseline_od,
                             "elevated", "normal")) %>%
    arrange(desc(od_first))

  baseline_per_sample <- baseline_check %>%
    filter(well_type == "sample") %>%
    group_by(sample_id) %>%
    summarise(n_elevated = sum(baseline == "elevated"), n = n(),
              median_baseline = median(od_first), .groups = "drop") %>%
    arrange(desc(n_elevated))

  list(baseline_check = baseline_check, baseline_per_sample = baseline_per_sample)
}

1.3.3 Standard curve and rate extraction

Fits the GSH standard curve, extracts each well’s rate via a sliding-window linear regression gated on R-squared, and summarizes the positive control.

## -----------------------------------------------------------------------
## fit_standard_curve()
## Purpose : Extract t=0 GSH standard readings, quantify signal drift over
##           the run, compute per-concentration replicate statistics, flag
##           outlier wells by deviation from their triplicate median, and
##           fit three candidate calibration lines (all wells, per-
##           concentration means, outlier-excluded). The outlier-excluded
##           fit is the calibration used downstream.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with std_outlier_od
## Outputs : list with standards_t0, standard_drift, standard_summary,
##           standards_flagged, fit_all_wells, fit_means, fit_no_outliers,
##           fit_comparison, std_slope, std_intercept, std_r2, std_nmol_max
## -----------------------------------------------------------------------
fit_standard_curve <- function(plate_long, assay_params) {
  standards_all <- plate_long %>%
    filter(well_type == "standard") %>%
    select(well, std_nmol, time_min, od, source)

  standards_t0 <- standards_all %>%
    filter(time_min == 0) %>%
    select(well, std_nmol, od, source) %>%
    arrange(std_nmol, well)

  standard_drift <- standards_all %>%
    group_by(well, std_nmol) %>%
    summarise(od_t0 = od[time_min == 0], od_t40 = od[time_min == max(time_min)],
              drift = od_t40 - od_t0, .groups = "drop") %>%
    arrange(std_nmol, well)

  standard_summary <- standards_t0 %>%
    group_by(std_nmol) %>%
    summarise(n = n(), mean_od = mean(od), sd_od = sd(od),
              se_od = sd(od) / sqrt(n()), cv_pct = 100 * sd(od) / mean(od),
              median_od = median(od), .groups = "drop") %>%
    arrange(std_nmol) %>%
    mutate(net_od = mean_od - mean_od[std_nmol == 0],
           od_per_nmol = ifelse(std_nmol > 0, net_od / std_nmol, NA_real_))

  standards_flagged <- standards_t0 %>%
    group_by(std_nmol) %>%
    mutate(triplicate_median = median(od),
           deviation = od - triplicate_median,
           is_outlier = abs(deviation) > assay_params$std_outlier_od) %>%
    ungroup() %>%
    arrange(desc(abs(deviation)))

  fit_all_wells   <- lm(od ~ std_nmol, data = standards_flagged)
  fit_means       <- lm(mean_od ~ std_nmol, data = standard_summary)
  fit_no_outliers <- lm(od ~ std_nmol, data = standards_flagged %>% filter(!is_outlier))

  fit_comparison <- bind_rows(
    data.frame(fit = "all wells",          n = nobs(fit_all_wells),
               slope = coef(fit_all_wells)[2],   intercept = coef(fit_all_wells)[1],
               r_squared = summary(fit_all_wells)$r.squared),
    data.frame(fit = "concentration means", n = nobs(fit_means),
               slope = coef(fit_means)[2],       intercept = coef(fit_means)[1],
               r_squared = summary(fit_means)$r.squared),
    data.frame(fit = "outlier-excluded",   n = nobs(fit_no_outliers),
               slope = coef(fit_no_outliers)[2], intercept = coef(fit_no_outliers)[1],
               r_squared = summary(fit_no_outliers)$r.squared)
  )

  list(
    standards_all      = standards_all,
    standards_t0       = standards_t0,
    standard_drift     = standard_drift,
    standard_summary   = standard_summary,
    standards_flagged  = standards_flagged,
    fit_all_wells      = fit_all_wells,
    fit_means          = fit_means,
    fit_no_outliers    = fit_no_outliers,
    fit_comparison     = fit_comparison,
    std_slope          = unname(coef(fit_no_outliers)[2]),
    std_intercept      = unname(coef(fit_no_outliers)[1]),
    std_r2             = summary(fit_no_outliers)$r.squared,
    std_nmol_max       = max(standard_summary$std_nmol)
  )
}

## -----------------------------------------------------------------------
## window_slopes()
## Purpose : Fit every sliding window of `w` consecutive timepoints on one
##           kinetic trace and return the slope (mOD/min) and R^2 of each.
## Inputs  : t - numeric vector of timepoints (min)
##           y - numeric vector of OD readings, same length as t
##           w - window width in number of points
## Outputs : data.frame, one row per window: t_start, t_end,
##           slope_mOD_min, r2
## -----------------------------------------------------------------------
window_slopes <- function(t, y, w) {
  n <- length(y)
  bind_rows(lapply(seq_len(n - w + 1), function(s) {
    i  <- s:(s + w - 1)
    f  <- lm(y[i] ~ t[i])
    r2 <- summary(f)$r.squared
    data.frame(t_start = t[i[1]], t_end = t[i[w]],
               slope_mOD_min = unname(coef(f)[2]) * 1000,
               r2 = ifelse(is.na(r2), 0, r2))
  }))
}

## -----------------------------------------------------------------------
## compute_well_rates()
## Purpose : For every well, scan all sliding windows of assay_params$
##           rate_window_n consecutive reads and record the max-increasing
##           window (used for activity) and the max-absolute window
##           (diagnostic only -- matches Gen5's own `Max V` convention,
##           which can report large negative "rates" for degrading wells).
##           A rate is `rate_usable` only if its R^2 clears the floor, the
##           well's net change is positive, and no read glitch falls inside
##           the fitted window.
## Inputs  : plate_long       - data.frame from annotate_wells()
##           well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with rate_window_n, rate_min_r2
## Outputs : well_rates - one row per well: t_start, t_end, slope_mOD_min,
##           r2, max_abs_slope_mOD_min, abs_window_is_negative,
##           glitch_in_window, rate_usable (plus diagnostic columns joined in)
## -----------------------------------------------------------------------
compute_well_rates <- function(plate_long, well_diagnostics, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, well_type, sample_id, std_nmol) %>%
    group_modify(~ {
      w   <- window_slopes(.x$time_min, .x$od, assay_params$rate_window_n)
      inc <- w[which.max(w$slope_mOD_min), ]
      abs_ <- w[which.max(abs(w$slope_mOD_min)), ]
      data.frame(
        t_start = inc$t_start, t_end = inc$t_end,
        slope_mOD_min = inc$slope_mOD_min, r2 = inc$r2,
        max_abs_slope_mOD_min = abs_$slope_mOD_min,
        abs_window_is_negative = abs_$slope_mOD_min < 0
      )
    }) %>%
    ungroup() %>%
    left_join(well_diagnostics %>% select(well, net_change, frac_rising, od_max,
                                          flag_decreasing, flag_over_range,
                                          flag_glitch, glitch_at_min,
                                          flag_bg_active, n_flags),
              by = "well") %>%
    mutate(
      glitch_in_window = flag_glitch & glitch_at_min > t_start & glitch_at_min <= t_end,
      rate_usable      = r2 >= assay_params$rate_min_r2 & net_change > 0 & !glitch_in_window
    )
}

## -----------------------------------------------------------------------
## compute_positive_control()
## Purpose : Compute the CS Positive Control's rate and activity, and check
##           that its replicates rose linearly -- the plate's proof that the
##           reaction chemistry (DTNB, coupling, reader) worked, independent
##           of any sample-specific problem.
## Inputs  : well_rates   - data.frame from compute_well_rates()
##           std_slope    - calibration slope (A412/nmol), from fit_standard_curve()
##           assay_params - list with sample_volume_uL, dilution_factor
## Outputs : pos_control - one row per positive-control well: t_start, t_end,
##           slope_mOD_min, r2, net_change, activity_mU_uL
## -----------------------------------------------------------------------
compute_positive_control <- function(well_rates, std_slope, assay_params) {
  well_rates %>%
    filter(well_type == "positive_control") %>%
    mutate(activity_mU_uL = (slope_mOD_min / 1000) / std_slope /
                            assay_params$sample_volume_uL * assay_params$dilution_factor)
}

1.3.4 Background correction, replicate CV, and activity calculation

Estimates a single pooled background rate from the flat (well-behaved) background replicates, computes technical-replicate coefficients of variation, and converts corrected rates into protein-normalized CS activity.

## -----------------------------------------------------------------------
## compute_background_correction()
## Purpose : Estimate a single, plate-wide background rate from the flat
##           (well-behaved) replicates of the pooled background control
##           (note 3), since anomalous background wells do not measure
##           background. Unlike earlier plates' per-sample background
##           design, this plate has one background triplicate for the whole
##           plate, so the estimate is a single row rather than one row per
##           sample_id.
## Inputs  : well_rates - data.frame from compute_well_rates()
## Outputs : background_correction - one-row data.frame: n_bg_total,
##           n_bg_flat, bg_rate_flat (NA if no flat replicate exists),
##           bg_rate_all
## -----------------------------------------------------------------------
compute_background_correction <- function(well_rates) {
  bg <- well_rates %>% filter(well_type == "background")
  data.frame(
    n_bg_total   = nrow(bg),
    n_bg_flat    = sum(!bg$flag_bg_active),
    bg_rate_flat = ifelse(sum(!bg$flag_bg_active) > 0,
                          mean(bg$slope_mOD_min[!bg$flag_bg_active]), NA_real_),
    bg_rate_all  = mean(bg$slope_mOD_min)
  )
}

## -----------------------------------------------------------------------
## compute_replicate_cv()
## Purpose : Compute technical-replicate coefficient of variation (CV%) on
##           the extracted rate, both across ALL three replicates and
##           across USABLE replicates only (rate_usable == TRUE). Flags any
##           sample whose CV exceeds assay_params$cv_threshold_pct.
## Inputs  : well_rates       - data.frame from compute_well_rates()
##           sample_id_order  - character vector giving the row order /
##                              complete sample-ID set to report (typically
##                              the plate's sample IDs, e.g. from
##                              protein_by_sample$sample_id)
##           assay_params     - list with cv_threshold_pct
## Outputs : cv_summary - one row per sample: n_all, mean_all, sd_all,
##           cv_all, n_usable, mean_usable, sd_usable, cv_usable,
##           excluded_wells, fails_cv_all, fails_cv_usable
## -----------------------------------------------------------------------
compute_replicate_cv <- function(well_rates, sample_id_order, assay_params) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  cv_all <- sample_rates %>%
    group_by(sample_id) %>%
    summarise(n_all = n(), mean_all = mean(slope_mOD_min), sd_all = sd(slope_mOD_min),
              cv_all = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  cv_usable <- sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_usable = n(), mean_usable = mean(slope_mOD_min),
              sd_usable = sd(slope_mOD_min),
              cv_usable = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  data.frame(sample_id = sample_id_order, stringsAsFactors = FALSE) %>%
    left_join(cv_all, by = "sample_id") %>%
    left_join(cv_usable, by = "sample_id") %>%
    mutate(
      excluded_wells = vapply(sample_id, function(s) {
        w <- sample_rates$well[sample_rates$sample_id == s & !sample_rates$rate_usable]
        if (length(w) == 0) "-" else paste(w, collapse = ", ")
      }, character(1)),
      fails_cv_all    = cv_all > assay_params$cv_threshold_pct,
      fails_cv_usable = !is.na(cv_usable) & cv_usable > assay_params$cv_threshold_pct
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## calculate_cs_activity()
## Purpose : Compute per-sample CS activity following Abcam sec 10.3, using
##           only usable replicates, corrected for the single pooled
##           background rate (note 3), scaled through the standard curve,
##           and normalized to total extracted protein (rather than tissue
##           weight).
## Inputs  : well_rates            - data.frame from compute_well_rates()
##           protein_by_sample     - data.frame from load_protein_concentrations()
##           background_correction - one-row data.frame from
##                                    compute_background_correction()
##           std_slope             - calibration slope (A412/nmol), from fit_standard_curve()
##           std_nmol_max          - max calibrated GSH concentration, from fit_standard_curve()
##           assay_params          - list with sample_volume_uL, dilution_factor,
##                                    homogenate_volume_uL, read_duration_min
##           plate_long            - data.frame from annotate_wells(), used to
##                                    attach family/individual/temperature
## Outputs : cs_activity - one row per sample: n_reps_used, mean_rate_mOD_min,
##           sd_rate, cv_rate, family, individual, temperature, conc_ug_mL,
##           total_protein_mg, bg_rate_mOD_min, corrected_mOD_min,
##           activity_mU_per_uL, total_mU_in_homogenate,
##           activity_mU_per_mg_protein, within_std_range
## -----------------------------------------------------------------------
calculate_cs_activity <- function(well_rates, protein_by_sample, background_correction,
                                   std_slope, std_nmol_max, assay_params, plate_long) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  bg_rate <- ifelse(is.na(background_correction$bg_rate_flat), 0, background_correction$bg_rate_flat)

  sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_reps_used = n(), mean_rate_mOD_min = mean(slope_mOD_min),
              sd_rate = sd(slope_mOD_min),
              cv_rate = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop") %>%
    left_join(plate_long %>% filter(well_type == "sample") %>%
                distinct(sample_id, family, individual, temperature),
              by = "sample_id") %>%
    left_join(protein_by_sample %>% select(sample_id, conc_ug_mL, total_protein_mg),
              by = "sample_id") %>%
    mutate(
      bg_rate_mOD_min       = bg_rate,
      corrected_mOD_min     = mean_rate_mOD_min - bg_rate_mOD_min,
      rate_OD_min           = corrected_mOD_min / 1000,
      nmol_per_min          = rate_OD_min / std_slope,
      activity_mU_per_uL    = nmol_per_min / assay_params$sample_volume_uL *
                              assay_params$dilution_factor,
      total_mU_in_homogenate = activity_mU_per_uL * assay_params$homogenate_volume_uL,
      activity_mU_per_mg_protein = total_mU_in_homogenate / total_protein_mg,
      nmol_in_window        = nmol_per_min * (assay_params$read_duration_min),
      within_std_range      = nmol_in_window <= std_nmol_max
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## build_results_table()
## Purpose : Assemble the final, presentation-ready per-sample results
##           table from cs_activity, cv_summary and baseline diagnostics,
##           with a plain-language Interpretation column.
## Inputs  : cs_activity          - data.frame from calculate_cs_activity()
##           cv_summary           - data.frame from compute_replicate_cv()
##           baseline_per_sample  - data.frame from compute_baseline_diagnostics()
##           assay_params         - list with cv_threshold_pct, homogenate_volume_uL
## Outputs : results_table - formatted data.frame, ready for kable()/write.csv()
## -----------------------------------------------------------------------
build_results_table <- function(cs_activity, cv_summary, baseline_per_sample, assay_params) {
  cs_activity %>%
    left_join(cv_summary %>% select(sample_id, n_all, cv_all), by = "sample_id") %>%
    left_join(baseline_per_sample %>% select(sample_id, n_elevated), by = "sample_id") %>%
    transmute(
      Sample                       = sample_id,
      Family                       = family,
      Individual                   = individual,
      Temperature                  = temperature,
      `Protein conc (ug/mL)`       = round(conc_ug_mL, 1),
      `Total protein (mg)`         = round(total_protein_mg, 3),
      `Reps used`                  = paste0(n_reps_used, "/", n_all),
      `CV all reps (%)`            = round(cv_all, 1),
      `CV used reps (%)`           = round(cv_rate, 1),
      `Rate (mA412/min)`           = round(mean_rate_mOD_min, 2),
      `BG rate (mA412/min)`        = round(bg_rate_mOD_min, 3),
      `Corrected rate (mA412/min)` = round(corrected_mOD_min, 2),
      `Activity (mU/uL)`           = round(activity_mU_per_uL, 4),
      `Activity (mU/mg protein)`   = round(activity_mU_per_mg_protein, 3),
      `CV flag`                    = ifelse(cv_all > assay_params$cv_threshold_pct,
                                            paste0("FAIL >", assay_params$cv_threshold_pct, "%"), "pass"),
      `Elevated baseline reps`     = paste0(n_elevated, "/3"),
      Interpretation               = case_when(
        n_elevated == 3 ~ "DO NOT USE - no clean replicate",
        n_elevated > 0  ~ "caution - some reps compromised",
        cv_rate > assay_params$cv_threshold_pct | is.na(cv_rate) ~ "caution - imprecise",
        TRUE ~ "usable"
      )
    )
}

1.3.5 QC summary and narrative generation

Assembles the per-well QC summary table and auto-generates the plain-language QC findings bullets used in the QC SUMMARY section below.

## -----------------------------------------------------------------------
## build_qc_summary_table()
## Purpose : Consolidate every QC check performed across the pipeline into a
##           single two-column (check, value) table, suitable for kable() and
##           for writing to qc_summary.csv. Mirrors the individual find
##           functions but as compact, at-a-glance figures rather than prose.
## Inputs  : layout_wells, reconcile_summary, std_curve, well_diagnostics,
##           well_rates, cv_summary, baseline_per_sample, pos_control,
##           assay_params
## Outputs : qc_summary - data.frame with columns check, value (both chr)
## -----------------------------------------------------------------------
build_qc_summary_table <- function(layout_wells, reconcile_summary, std_curve,
                                    well_diagnostics, well_rates, cv_summary,
                                    baseline_per_sample, pos_control, assay_params) {
  standard_summary  <- std_curve$standard_summary
  standards_flagged <- std_curve$standards_flagged
  standard_drift    <- std_curve$standard_drift
  sample_rates      <- well_rates %>% filter(well_type == "sample")

  data.frame(
    check = c(
      "Occupied wells in layout",
      "Layout wells missing from absorbance CSV",
      "Max disagreement between absorbance CSV and full report",
      "Standard concentrations with replicate CV > threshold",
      "Standard wells flagged as outliers",
      "Standard curve R^2 (outlier-excluded, replicate level)",
      "Standard curve R^2 (all wells, replicate level)",
      "Standard wells that LOST signal over the run",
      "Standards above photometric linearity ceiling",
      "Background control wells behaving as active reactions",
      "Sample wells with a decreasing trace",
      "Sample wells with an elevated starting A412",
      "Samples with ALL THREE replicates baseline-compromised",
      "Sample wells with a read glitch",
      "Sample wells usable for rate extraction",
      "Samples with technical CV > threshold (all reps)",
      "Samples with technical CV > threshold (usable reps)",
      "Positive control replicates rising and linear"
    ),
    value = c(
      nrow(layout_wells),
      reconcile_summary$n_wells_layout_not_csv,
      sprintf("%.g", reconcile_summary$max_disagreement),
      paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
            collapse = ", "),
      paste0(sum(standards_flagged$is_outlier), "/", nrow(standards_flagged), " (",
             paste(standards_flagged$well[standards_flagged$is_outlier], collapse = ", "), ")"),
      sprintf("%.4f", std_curve$std_r2),
      sprintf("%.4f", summary(std_curve$fit_all_wells)$r.squared),
      paste0(sum(standard_drift$drift < 0), "/", nrow(standard_drift)),
      paste(standard_summary$std_nmol[standard_summary$mean_od > assay_params$od_linear_max],
            collapse = ", "),
      paste0(sum(well_diagnostics$flag_bg_active), "/",
             sum(well_diagnostics$well_type == "background")),
      paste0(sum(well_diagnostics$flag_decreasing & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(well_diagnostics$flag_high_baseline), "/",
             sum(well_diagnostics$well_type == "sample"), " (",
             paste(well_diagnostics$well[well_diagnostics$flag_high_baseline], collapse = ", "), ")"),
      paste0(sum(baseline_per_sample$n_elevated == baseline_per_sample$n), "/",
             nrow(baseline_per_sample), " (",
             paste(baseline_per_sample$sample_id[baseline_per_sample$n_elevated == baseline_per_sample$n],
                   collapse = ", "), ")"),
      paste0(sum(well_diagnostics$flag_glitch & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(sample_rates$rate_usable), "/", nrow(sample_rates)),
      paste0(sum(cv_summary$fails_cv_all), "/", nrow(cv_summary)),
      paste0(sum(cv_summary$fails_cv_usable), "/", nrow(cv_summary)),
      paste0(sum(pos_control$net_change > 0 & pos_control$r2 > 0.99), "/", nrow(pos_control))
    ),
    stringsAsFactors = FALSE
  )
}

## -----------------------------------------------------------------------
## generate_qc_findings()
## Purpose : Walk the QC/CV/anomaly objects for a plate and emit a
##           consistent, generic-language markdown bullet list of findings
##           (decreasing/glitched/high-baseline wells, background-control
##           mix-ups, standard-curve QC failures, CV failures, disqualified
##           samples). Intended to seed the SUMMARY section of a plate's
##           .Rmd; a short hand-written subsection for plate-specific
##           interpretation should follow it.
## Inputs  : well_diagnostics    - data.frame from compute_well_diagnostics()
##           well_rates          - data.frame from compute_well_rates()
##           cv_summary          - data.frame from compute_replicate_cv()
##           baseline_per_sample - data.frame from compute_baseline_diagnostics()
##           std_curve           - list from fit_standard_curve()
##           reconcile_summary   - list from reconcile_kinetic_sources()$summary
##           assay_params        - list with cv_threshold_pct, od_linear_max
## Outputs : character vector, one markdown bullet per finding (empty
##           findings are omitted, so a clean plate returns a short list)
## -----------------------------------------------------------------------
generate_qc_findings <- function(well_diagnostics, well_rates, cv_summary,
                                  baseline_per_sample, std_curve,
                                  reconcile_summary, assay_params) {
  bullets <- character(0)

  # Source reconciliation
  if (reconcile_summary$n_wells_layout_not_csv > 0 ||
      (!is.na(reconcile_summary$max_disagreement) && reconcile_summary$max_disagreement >= 1e-9)) {
    bullets <- c(bullets, sprintf(
      "**Kinetic source disagreement.** %d layout well(s) missing from the absorbance CSV; max disagreement between absorbance CSV and full report was %.4g A412.",
      reconcile_summary$n_wells_layout_not_csv, reconcile_summary$max_disagreement))
  } else {
    bullets <- c(bullets, sprintf(
      "**Kinetic sources agree.** absorbance CSV and full report agree exactly on all %d shared well x timepoint readings; all layout wells are present.",
      reconcile_summary$n_shared_readings))
  }

  # Sample wells with decreasing kinetics
  dec_samples <- well_diagnostics %>% filter(well_type == "sample", flag_decreasing)
  if (nrow(dec_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Decreasing kinetics.** %d sample well(s) end lower than (or equal to) their starting A412 and are excluded from rate extraction: %s.",
      nrow(dec_samples), paste(dec_samples$well, collapse = ", ")))
  }

  # Read glitches
  glitch_samples <- well_diagnostics %>% filter(well_type == "sample", flag_glitch)
  if (nrow(glitch_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Read glitches.** %d sample well(s) contain a single read-to-read step inconsistent with its neighbours: %s (only disqualifying if it falls inside the fitted rate window).",
      nrow(glitch_samples), paste(glitch_samples$well, collapse = ", ")))
  }

  # Elevated baseline
  high_baseline <- well_diagnostics %>% filter(well_type == "sample", flag_high_baseline)
  if (nrow(high_baseline) > 0) {
    bullets <- c(bullets, sprintf(
      "**Elevated starting absorbance.** %d sample well(s) start above the baseline threshold (A412 > %s), indicating pre-existing thiol/contamination before the reaction began: %s.",
      nrow(high_baseline), assay_params$sample_baseline_od, paste(high_baseline$well, collapse = ", ")))
  }

  # Samples fully disqualified by baseline
  disqualified <- baseline_per_sample %>% filter(n_elevated == n)
  if (nrow(disqualified) > 0) {
    bullets <- c(bullets, sprintf(
      "**Sample(s) with no clean replicate.** Every replicate of %s has an elevated starting A412; these samples have no usable measurement on this plate regardless of CV.",
      paste(disqualified$sample_id, collapse = ", ")))
  }

  # Over-range wells
  over_range <- well_diagnostics %>% filter(flag_over_range)
  if (nrow(over_range) > 0) {
    bullets <- c(bullets, sprintf(
      "**Over-range wells.** %d well(s) exceed the photometric linearity ceiling (A412 > %s): %s.",
      nrow(over_range), assay_params$od_linear_max, paste(over_range$well, collapse = ", ")))
  }

  # Background wells behaving as active reactions
  bg_active <- well_diagnostics %>% filter(well_type == "background", flag_bg_active)
  if (nrow(bg_active) > 0) {
    bullets <- c(bullets, sprintf(
      "**Background control wells behaving as active reactions.** %d of %d background wells show a starting A412 or drift inconsistent with a flat, inactive well: %s.",
      nrow(bg_active), sum(well_diagnostics$well_type == "background"),
      paste(bg_active$well, collapse = ", ")))
  }

  # Standard curve QC
  std_summary <- std_curve$standard_summary
  cv_fail_std <- std_summary %>% filter(cv_pct > assay_params$cv_threshold_pct)
  if (nrow(cv_fail_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**GSH standard curve replicate imprecision.** %d of %d concentrations exceed %s%% CV: %s nmol/well.",
      nrow(cv_fail_std), nrow(std_summary), assay_params$cv_threshold_pct,
      paste(cv_fail_std$std_nmol, collapse = ", ")))
  }
  n_std_outliers <- sum(std_curve$standards_flagged$is_outlier)
  if (n_std_outliers > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard outlier wells.** %d of %d standard wells deviate from their triplicate median by more than %s A412: %s.",
      n_std_outliers, nrow(std_curve$standards_flagged), assay_params$std_outlier_od,
      paste(std_curve$standards_flagged$well[std_curve$standards_flagged$is_outlier], collapse = ", ")))
  }
  over_range_std <- std_summary %>% filter(mean_od > assay_params$od_linear_max)
  if (nrow(over_range_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**Standards above the photometric linear range.** %s nmol/well read above A412 %s; the calibration is trustworthy only below these concentrations.",
      paste(over_range_std$std_nmol, collapse = ", "), assay_params$od_linear_max))
  }
  n_drift_neg <- sum(std_curve$standard_drift$drift < 0)
  if (n_drift_neg > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard signal decay.** %d of %d standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.",
      n_drift_neg, nrow(std_curve$standard_drift)))
  }
  bullets <- c(bullets, sprintf(
    "**Standard curve fit (outlier-excluded):** slope = %.5f A412/nmol, R^2 = %.4f (all-wells R^2 = %.4f).",
    std_curve$std_slope, std_curve$std_r2, summary(std_curve$fit_all_wells)$r.squared))

  # Technical replicate CV failures
  fail_all <- cv_summary %>% filter(fails_cv_all)
  if (nrow(fail_all) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% (all replicates).** %d of %d samples: %s.",
      assay_params$cv_threshold_pct, nrow(fail_all), nrow(cv_summary),
      paste(sprintf("%s (%.0f%%)", fail_all$sample_id, fail_all$cv_all), collapse = ", ")))
  } else {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV.** All %d samples are within %s%% CV across all three replicates.",
      nrow(cv_summary), assay_params$cv_threshold_pct))
  }
  fail_usable <- cv_summary %>% filter(fails_cv_usable)
  if (nrow(fail_usable) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% persists after excluding flagged replicates.** %s.",
      assay_params$cv_threshold_pct,
      paste(sprintf("%s (%.0f%%, n=%d)", fail_usable$sample_id, fail_usable$cv_usable,
                    fail_usable$n_usable), collapse = ", ")))
  }

  bullets
}

1.4 Assay parameters

All assay constants and QC thresholds are collected here so nothing downstream hard-codes a number.

assay_params <- list(
  sample_volume_uL      = 2,     # homogenate volume per reaction well (V in Abcam formula); also the pooled-background volume (note 3)
  buffer_volume_uL      = 48,    # Assay Buffer 7 added to reach 50 uL pre-Reaction Mix
  dilution_factor       = 1,     # D; layout `df.0` = undiluted homogenate
  homogenate_volume_uL  = 350,   # MEASURED buffer volume tissue was homogenized in (see note 2)
  read_wavelength_nm    = 412,
  read_interval_min     = 2,
  read_duration_min     = 20,    # see note 1
  rate_window_n         = 5,     # points per sliding regression window (5 pts = 8 min)
  rate_min_r2           = 0.80,  # min R^2 for a rate window to be trusted
  glitch_excess_od      = 0.02,  # |step - mean of neighbouring steps| flagging a read glitch
  cv_threshold_pct      = 15,    # technical-replicate CV QC threshold
  od_linear_max         = 1.5,   # upper A412 bound for reliable photometry
  std_outlier_od        = 0.15,  # |deviation from triplicate median| flagging a standard well
  sample_baseline_od    = 0.35,  # max acceptable t0 A412 for a sample well
  bg_flat_od_max        = 0.15,  # a well-behaved background well starts below this A412
  bg_flat_drift_max     = 0.05   # ...and drifts less than this over the whole run
)

cat("--- assay_params: assay constants and QC thresholds ---\n\n")
str(assay_params)
--- assay_params: assay constants and QC thresholds ---

List of 16
 $ sample_volume_uL    : num 2
 $ buffer_volume_uL    : num 48
 $ dilution_factor     : num 1
 $ homogenate_volume_uL: num 350
 $ read_wavelength_nm  : num 412
 $ read_interval_min   : num 2
 $ read_duration_min   : num 20
 $ rate_window_n       : num 5
 $ rate_min_r2         : num 0.8
 $ glitch_excess_od    : num 0.02
 $ cv_threshold_pct    : num 15
 $ od_linear_max       : num 1.5
 $ std_outlier_od      : num 0.15
 $ sample_baseline_od  : num 0.35
 $ bg_flat_od_max      : num 0.15
 $ bg_flat_drift_max   : num 0.05

1.5 Output directory

# Output directory for this analysis (matches this file's name, per ../code/README.md)
output_dir <- "../outputs/Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- output_dir: destination for all figures and tables ---\n")
str(output_dir)
--- output_dir: destination for all figures and tables ---
 chr "../outputs/Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient"

2 DATA

Data are read from the local repo (../data/raw_absorbance/) so this document renders before/after the files are pushed to GitHub.

2.1 Plate layout

data_dir <- "../data/raw_absorbance"
run_stem <- "Gen5-20260824-mgig-sormi-citrate_synthase-F05-ambient"

layout_wells <- parse_plate_layout(file.path(data_dir, paste0("layout-", run_stem, ".csv")))

cat("Layout format detected:", attr(layout_wells, "layout_format"), "\n")
cat("Occupied wells in layout:", nrow(layout_wells), "\n")

cat("\n--- layout_wells: one row per occupied well, with its descriptive label ---\n\n")
str(layout_wells)
Layout format detected: single 
Occupied wells in layout: 48 

--- layout_wells: one row per occupied well, with its descriptive label ---

'data.frame':   48 obs. of  4 variables:
 $ well     : chr  "A1" "A2" "A3" "A4" ...
 $ plate_row: chr  "A" "A" "A" "A" ...
 $ plate_col: int  1 2 3 4 5 6 7 8 9 10 ...
 $ label    : chr  "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_02_ambient-citrate_synthase-17.5-df.0" ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "row_idx" "col_idx"
  ..$ dimnames:List of 2
  .. ..$ row_idx: chr [1:8] "row_idx=2" "row_idx=3" "row_idx=4" "row_idx=5" ...
  .. ..$ col_idx: chr [1:12] "col_idx= 2" "col_idx= 3" "col_idx= 4" "col_idx= 5" ...
 - attr(*, "layout_format")= chr "single"

2.2 Kinetic readers

absorbance_csv <- read_absorbance_csv(file.path(data_dir, paste0("absorbance-", run_stem, ".csv")))
full_report    <- read_full_report(file.path(data_dir, paste0("full_report-", run_stem, ".txt")))

cat("absorbance CSV : ", dplyr::n_distinct(absorbance_csv$well), " wells x ",
    dplyr::n_distinct(absorbance_csv$time_min), " timepoints\n", sep = "")
cat("full report    : ", dplyr::n_distinct(full_report$well), " wells x ",
    dplyr::n_distinct(full_report$time_min), " timepoints\n", sep = "")

cat("\n--- absorbance_csv: long-format readings from absorbance-*.csv ---\n\n")
str(absorbance_csv)
cat("\n--- full_report: long-format readings from full_report-*.txt ---\n\n")
str(full_report)
absorbance CSV : 48 wells x 11 timepoints
full report    : 48 wells x 11 timepoints

--- absorbance_csv: long-format readings from absorbance-*.csv ---

'data.frame':   528 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.135 0.15 0.165 0.181 0.196 0.212 0.227 0.242 0.255 0.269 ...

--- full_report: long-format readings from full_report-*.txt ---

'data.frame':   528 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.135 0.15 0.165 0.181 0.196 0.212 0.227 0.242 0.255 0.269 ...

2.3 Cross-check against the full report

absorbance-*.csv is used for every well; full_report-*.txt is used only to confirm the CSV is trustworthy: every well it reports should also be in the CSV, every well the layout expects should be in the CSV, and every reading the two files share should agree exactly.

recon <- reconcile_kinetic_sources(absorbance_csv, full_report, layout_wells)
plate_readings <- recon$plate_readings

cat("Wells in absorbance CSV          :", recon$summary$n_wells_csv, "\n")
cat("Wells in full report             :", recon$summary$n_wells_report, "\n")
cat("Wells in full report not in CSV  :", recon$summary$n_wells_report_not_csv, "\n")
cat("Layout wells missing from CSV    :", recon$summary$n_wells_layout_not_csv, "\n")
cat("CSV wells not in layout          :", recon$summary$n_wells_csv_not_layout, "\n")
cat("Shared well x timepoints         :", recon$summary$n_shared_readings, "\n")
cat("Max |CSV - report| disagreement  :", recon$summary$max_disagreement, "\n")

cat("\n--- recon$overlap_check: per-reading comparison of the two raw files ---\n\n")
str(recon$overlap_check)
cat("\n--- plate_readings: readings for all occupied wells, from absorbance_csv ---\n\n")
str(plate_readings)
Wells in absorbance CSV          : 48 
Wells in full report             : 48 
Wells in full report not in CSV  : 0 
Layout wells missing from CSV    : 0 
CSV wells not in layout          : 0 
Shared well x timepoints         : 528 
Max |CSV - report| disagreement  : 0 

--- recon$overlap_check: per-reading comparison of the two raw files ---

'data.frame':   528 obs. of  5 variables:
 $ well     : chr  "A1" "A1" "A1" "A1" ...
 $ time_min : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od_csv   : num  0.135 0.15 0.165 0.181 0.196 0.212 0.227 0.242 0.255 0.269 ...
 $ od_report: num  0.135 0.15 0.165 0.181 0.196 0.212 0.227 0.242 0.255 0.269 ...
 $ abs_diff : num  0 0 0 0 0 0 0 0 0 0 ...

--- plate_readings: readings for all occupied wells, from absorbance_csv ---

'data.frame':   528 obs. of  4 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.135 0.15 0.165 0.181 0.196 0.212 0.227 0.242 0.255 0.269 ...
 $ source  : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...

This plate’s absorbance-*.csv covers the full 48 occupied wells (24 sample + 18 standard + 3 positive-control + 3 pooled-background, note 3) and the layout carries no spelling typo, so no label fix is needed here.

2.4 Annotate wells and parse metadata

plate_long <- suppressWarnings(annotate_wells(plate_readings, layout_wells))

cat("Wells per type:\n")
print(plate_long %>% distinct(well, well_type) %>% count(well_type, name = "n_wells") %>%
        as.data.frame(), row.names = FALSE)
cat("\nSamples:", paste(sort(unique(na.omit(plate_long$sample_id))), collapse = ", "), "\n")
cat("Standards (nmol/well):", paste(sort(unique(na.omit(plate_long$std_nmol))), collapse = ", "), "\n")

cat("\n--- plate_long: fully annotated long-format plate, one row per well x timepoint ---\n\n")
str(plate_long)
Wells per type:
        well_type n_wells
       background       3
 positive_control       3
           sample      24
         standard      18

Samples: F05_01_ambient, F05_02_ambient, F05_03_ambient, F05_04_ambient, F05_05_ambient, F05_06_ambient, F05_07_ambient, F05_08_ambient 
Standards (nmol/well): 0, 8, 16, 24, 32, 40 

--- plate_long: fully annotated long-format plate, one row per well x timepoint ---

'data.frame':   528 obs. of  15 variables:
 $ well       : chr  "D10" "D10" "D10" "D10" ...
 $ time_min   : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od         : num  0.064 0.064 0.064 0.065 0.065 0.065 0.066 0.066 0.067 0.067 ...
 $ source     : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ plate_row  : chr  "D" "D" "D" "D" ...
 $ plate_col  : int  10 10 10 10 10 10 10 10 10 10 ...
 $ label      : chr  "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" ...
 $ well_type  : chr  "background" "background" "background" "background" ...
 $ label_clean: chr  "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" ...
 $ sample_id  : chr  NA NA NA NA ...
 $ family     : chr  NA NA NA NA ...
 $ individual : chr  NA NA NA NA ...
 $ temperature: chr  NA NA NA NA ...
 $ weight_mg  : num  NA NA NA NA NA NA NA NA NA NA ...
 $ std_nmol   : num  NA NA NA NA NA NA NA NA NA NA ...

annotate_wells() emits a harmless NAs introduced by coercion warning (suppressed above), because its weight_mg parser only applies to sample wells and returns NA for standards/positive control/background by construction (background is a pooled control with no individual weight of its own, note 3). No weight_mg value is actually missing for a sample well; annotate_wells()’s internal stopifnot() would halt otherwise.

2.5 Protein concentration (normalization factor)

Activity is normalized to total extracted protein rather than tissue weight (note 2). Protein concentration for each sample was measured on a separate BCA/Bradford-style plate (595 nm) and supplied as two CSV exports – the same measurements used for the original 2026-08-11 assay of these homogenates. No Sample string typo needed correction for this plate’s samples.

protein_files <- c(
  "../data/BSA/raw_absorbance/sample_protein_concentrations_1.csv",
  "../data/BSA/raw_absorbance/sample_protein_concentrations_2.csv"
)

plate_sample_ids <- sort(unique(na.omit(plate_long$sample_id)))

protein_by_sample <- load_protein_concentrations(protein_files, plate_sample_ids,
                                                  assay_params$homogenate_volume_uL) %>%
  left_join(plate_long %>% filter(well_type == "sample") %>%
              distinct(sample_id, family, individual, temperature),
            by = "sample_id")

cat("Samples matched to a protein concentration record:", nrow(protein_by_sample),
    "/", length(plate_sample_ids), "\n")
cat("Protein concentration range (ug/mL):",
    paste(range(protein_by_sample$conc_ug_mL), collapse = " - "), "\n")
cat("Total protein per homogenate range (mg):",
    paste(round(range(protein_by_sample$total_protein_mg), 3), collapse = " - "), "\n\n")
print(protein_by_sample %>% select(sample_id, source_file, conc_ug_mL, total_protein_mg) %>%
        as.data.frame(), row.names = FALSE)

cat("\n--- protein_by_sample: matched protein concentration and total protein per homogenate ---\n\n")
str(protein_by_sample)
Samples matched to a protein concentration record: 8 / 8 
Protein concentration range (ug/mL): 711.7 - 2703.7 
Total protein per homogenate range (mg): 0.249 - 0.946 

      sample_id                         source_file conc_ug_mL total_protein_mg
 F05_01_ambient sample_protein_concentrations_2.csv      876.1         0.306635
 F05_02_ambient sample_protein_concentrations_2.csv     1038.9         0.363615
 F05_03_ambient sample_protein_concentrations_2.csv     1900.4         0.665140
 F05_04_ambient sample_protein_concentrations_1.csv     1129.4         0.395290
 F05_05_ambient sample_protein_concentrations_1.csv      711.7         0.249095
 F05_06_ambient sample_protein_concentrations_2.csv     2232.4         0.781340
 F05_07_ambient sample_protein_concentrations_1.csv     1687.8         0.590730
 F05_08_ambient sample_protein_concentrations_2.csv     2703.7         0.946295

--- protein_by_sample: matched protein concentration and total protein per homogenate ---

'data.frame':   8 obs. of  7 variables:
 $ sample_id       : chr  "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ source_file     : chr  "sample_protein_concentrations_2.csv" "sample_protein_concentrations_2.csv" "sample_protein_concentrations_2.csv" "sample_protein_concentrations_1.csv" ...
 $ conc_ug_mL      : num  876 1039 1900 1129 712 ...
 $ total_protein_mg: num  0.307 0.364 0.665 0.395 0.249 ...
 $ family          : chr  "F05" "F05" "F05" "F05" ...
 $ individual      : chr  "01" "02" "03" "04" ...
 $ temperature     : chr  "ambient" "ambient" "ambient" "ambient" ...

3 KINETIC TRACES

Every trace is plotted before any rate is extracted. In this assay the shape of the trace is the primary diagnostic: CS activity must produce a monotonically rising A412, and only a genuinely linear stretch may be used for a rate.

trace_plot <- ggplot(plate_long, aes(x = time_min, y = od, group = well, colour = well_type)) +
  geom_line(linewidth = 0.5) +
  geom_hline(yintercept = assay_params$od_linear_max,
             linetype = "dotted", colour = "grey40") +
  facet_wrap(~ plate_row + plate_col, ncol = 12,
             labeller = function(d) list(paste0(d$plate_row, d$plate_col))) +
  scale_colour_manual(values = c(sample = "#08519c", background = "#a63603",
                                 standard = "#238b45", positive_control = "#6a51a3")) +
  labs(title = "A412 kinetic traces, all occupied wells",
       subtitle = paste0("Dotted line = assumed photometric linearity ceiling (A412 = ",
                         assay_params$od_linear_max,
                         "). Rising traces only are usable for rate extraction."),
       x = "Time (min)", y = "A412", colour = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        strip.text = element_text(size = 6),
        axis.text  = element_text(size = 5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "kinetic_traces_all_wells.png"), trace_plot,
       width = 11, height = 8, dpi = 300)

cat("--- trace_plot: ggplot object structure ---\n\n")
summary(trace_plot)

trace_plot

--- trace_plot: ggplot object structure ---

data: well, time_min, od, source, plate_row, plate_col, label,
  well_type, label_clean, sample_id, family, individual, temperature,
  weight_mg, std_nmol [528x15]
mapping:  x = ~time_min, y = ~od, group = ~well, colour = ~well_type
scales:   colour 
faceting:  ~plate_row, ~plate_col 
-----------------------------------
geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

Every sample and positive-control trace on this plate rises across the full 20-minute window (replicate-level R2 > 0.99 for the fitted rate window on every sample well, computed below). The three pooled-background traces (D10-D12) are visibly near-flat and far lower in absolute A412 than any sample well, consistent with a well-behaved background control. This plate shows no elevated-starting-absorbance or contamination signature in any sample well.

4 ANOMALY DETECTION

4.1 Per-well trace diagnostics

Four independent checks, computed from the raw trace before any rate fitting:

  1. Direction — net A412 change from first to last read. A CS-containing well must be positive.
  2. Monotonicity — fraction of the 10 read-to-read intervals that rise.
  3. Over-range — any read above the photometric linearity ceiling.
  4. Discontinuity — a single read-to-read step that departs from its two neighbouring steps, catching read glitches rather than biology.

Two further checks specific to well role: an elevated starting A412 (pre-existing thiol/contamination) for sample wells, and a background well whose trace is not flat (indicating it received Reaction Mix rather than Background Control Mix).

well_diagnostics <- compute_well_diagnostics(plate_long, assay_params)

cat("Flag counts across all", nrow(well_diagnostics), "wells:\n")
print(well_diagnostics %>% summarise(across(starts_with("flag_"), sum)) %>%
        as.data.frame(), row.names = FALSE)

cat("\nWells with at least one flag:", sum(well_diagnostics$n_flags > 0), "/", nrow(well_diagnostics), "\n")

cat("\n--- well_diagnostics: per-well trace diagnostics and anomaly flags ---\n\n")
str(well_diagnostics)
Flag counts across all 48 wells:
 flag_decreasing flag_over_range flag_glitch flag_high_baseline flag_bg_active
              15               0           0                  0              0

Wells with at least one flag: 15 / 48 

--- well_diagnostics: per-well trace diagnostics and anomaly flags ---

tibble [48 × 24] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:48] "A1" "A10" "A11" "A12" ...
 $ plate_row         : chr [1:48] "A" "A" "A" "A" ...
 $ plate_col         : int [1:48] 1 10 11 12 2 3 4 5 6 7 ...
 $ well_type         : chr [1:48] "sample" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:48] "F05_01_ambient" "F05_04_ambient" "F05_04_ambient" "F05_04_ambient" ...
 $ label             : chr [1:48] "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_04_ambient-citrate_synthase-14.0-df.0" "F05_04_ambient-citrate_synthase-14.0-df.0" "F05_04_ambient-citrate_synthase-14.0-df.0" ...
 $ std_nmol          : num [1:48] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:48] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:48] 0.135 0.121 0.12 0.122 0.12 0.119 0.119 0.124 0.119 0.128 ...
 $ od_last           : num [1:48] 0.28 0.293 0.297 0.301 0.268 0.257 0.274 0.285 0.271 0.314 ...
 $ od_max            : num [1:48] 0.28 0.293 0.297 0.301 0.268 0.257 0.274 0.285 0.271 0.314 ...
 $ net_change        : num [1:48] 0.145 0.172 0.177 0.179 0.148 0.138 0.155 0.161 0.152 0.186 ...
 $ frac_rising       : num [1:48] 1 1 1 1 1 1 1 1 1 1 ...
 $ max_step          : num [1:48] 0.016 0.02 0.021 0.021 0.016 ...
 $ typical_step      : num [1:48] 0.015 0.0185 0.0195 0.0195 0.015 ...
 $ step_excess       : num [1:48] -0.003 -0.001 -0.002 -0.003 -0.001 ...
 $ glitch_at_min     : num [1:48] 20 20 20 20 2 2 20 20 2 14 ...
 $ step_ratio        : num [1:48] 1.07 1.08 1.08 1.08 1.07 ...
 $ flag_decreasing   : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range   : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:48] 0 0 0 0 0 0 0 0 0 0 ...

4.2 Anomaly table

anomaly_table <- well_diagnostics %>%
  filter(n_flags > 0) %>%
  select(well, sample_id, well_type, od_first, net_change, n_flags,
         flag_decreasing, flag_over_range, flag_glitch,
         flag_high_baseline, flag_bg_active) %>%
  arrange(desc(n_flags), well)

kable(anomaly_table, digits = 3,
      caption = "Wells with at least one anomaly flag")

cat("\nTotal flagged wells:", nrow(anomaly_table), "/", nrow(well_diagnostics), "\n")

cat("\n--- anomaly_table: all flagged wells with their specific flags ---\n\n")
str(anomaly_table)
well sample_id well_type od_first net_change n_flags flag_decreasing flag_over_range flag_glitch flag_high_baseline flag_bg_active
C10 NA standard 0.869 -0.012 1 TRUE FALSE FALSE FALSE FALSE
C11 NA standard 0.862 -0.013 1 TRUE FALSE FALSE FALSE FALSE
C12 NA standard 0.865 -0.010 1 TRUE FALSE FALSE FALSE FALSE
C4 NA standard 0.367 -0.001 1 TRUE FALSE FALSE FALSE FALSE
C5 NA standard 0.381 0.000 1 TRUE FALSE FALSE FALSE FALSE
C6 NA standard 0.380 -0.001 1 TRUE FALSE FALSE FALSE FALSE
C7 NA standard 0.634 -0.004 1 TRUE FALSE FALSE FALSE FALSE
C8 NA standard 0.634 0.000 1 TRUE FALSE FALSE FALSE FALSE
C9 NA standard 0.619 -0.001 1 TRUE FALSE FALSE FALSE FALSE
D1 NA standard 1.126 -0.019 1 TRUE FALSE FALSE FALSE FALSE
D2 NA standard 1.104 -0.014 1 TRUE FALSE FALSE FALSE FALSE
D3 NA standard 1.123 -0.016 1 TRUE FALSE FALSE FALSE FALSE
D4 NA standard 1.419 -0.025 1 TRUE FALSE FALSE FALSE FALSE
D5 NA standard 1.430 -0.030 1 TRUE FALSE FALSE FALSE FALSE
D6 NA standard 1.433 -0.029 1 TRUE FALSE FALSE FALSE FALSE

Wells with at least one anomaly flag

Total flagged wells: 15 / 48 

--- anomaly_table: all flagged wells with their specific flags ---

tibble [15 × 11] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:15] "C10" "C11" "C12" "C4" ...
 $ sample_id         : chr [1:15] NA NA NA NA ...
 $ well_type         : chr [1:15] "standard" "standard" "standard" "standard" ...
 $ od_first          : num [1:15] 0.869 0.862 0.865 0.367 0.381 ...
 $ net_change        : num [1:15] -0.012 -0.013 -0.01 -0.001 0 ...
 $ n_flags           : int [1:15] 1 1 1 1 1 1 1 1 1 1 ...
 $ flag_decreasing   : logi [1:15] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ flag_over_range   : logi [1:15] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:15] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:15] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:15] FALSE FALSE FALSE FALSE FALSE FALSE ...

Every flagged well on this plate is a GSH standard, and every flag is flag_decreasing – 15 of the 18 standard wells (all five non-zero concentrations: 8, 16, 24, 32, 40 nmol) end at or below their own starting A412 over the 20-minute window, consistent with TNB2- signal decay (see GSH STANDARD CURVE below). This is a substantially larger fraction than the 20260814 plate (3/18 standards lost signal there), but because the standard curve is calibrated at t = 0 (not from the endpoint), it does not compromise the calibration (R2 = 0.9983, no outlier wells). No sample well, positive-control well, or background well is flagged for anything on this plate – no read glitches, no elevated baseline, and no background well behaving as an active reaction.

4.3 Background wells behaving as active reactions

bg_active_wells <- well_diagnostics %>%
  filter(well_type == "background", flag_bg_active) %>%
  select(well, sample_id, od_first, net_change)

kable(bg_active_wells, digits = 3,
      caption = "Background wells with a non-flat trace")

cat("Background wells behaving as active reactions:", nrow(bg_active_wells),
    "/", sum(well_diagnostics$well_type == "background"), "\n")

cat("\n--- bg_active_wells: background wells excluded from the background-rate estimate ---\n\n")
str(bg_active_wells)
well sample_id od_first net_change —— ———– ———- ————

Background wells with a non-flat trace

Background wells behaving as active reactions: 0 / 3

— bg_active_wells: background wells excluded from the background-rate estimate —

tibble [0 × 4] (S3: tbl_df/tbl/data.frame) $ well : chr(0) $ sample_id : chr(0) $ od_first : num(0) $ net_change: num(0)

All three pooled-background replicates (D10, D11, D12) are flat and well-behaved: starting A412 0.055-0.064 (well below the 0.15 threshold) and net change 0.002-0.003 over the full 20 minutes (well below the 0.05 drift threshold). All three are used in the background-rate estimate below.

4.4 Baseline diagnostic (elevated starting absorbance)

bd <- compute_baseline_diagnostics(well_diagnostics, assay_params)
baseline_check       <- bd$baseline_check
baseline_per_sample  <- bd$baseline_per_sample

cat("Sample/background wells with elevated starting A412 (> ",
    assay_params$sample_baseline_od, "):",
    sum(baseline_check$baseline == "elevated"), "/", nrow(baseline_check), "\n", sep = "")

kable(baseline_per_sample, digits = 3,
      caption = "Elevated-baseline replicate count per sample")

cat("\n--- baseline_check: per-well baseline classification ---\n\n")
str(baseline_check)
cat("\n--- baseline_per_sample: elevated-replicate count per sample ---\n\n")
str(baseline_per_sample)
Sample/background wells with elevated starting A412 (> 0.35):0/27
sample_id n_elevated n median_baseline
F05_01_ambient 0 3 0.120
F05_02_ambient 0 3 0.119
F05_03_ambient 0 3 0.128
F05_04_ambient 0 3 0.121
F05_05_ambient 0 3 0.119
F05_06_ambient 0 3 0.133
F05_07_ambient 0 3 0.124
F05_08_ambient 0 3 0.131

Elevated-baseline replicate count per sample

--- baseline_check: per-well baseline classification ---

tibble [27 × 25] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:27] "B4" "A1" "B10" "B5" ...
 $ plate_row         : chr [1:27] "B" "A" "B" "B" ...
 $ plate_col         : int [1:27] 4 1 10 5 11 6 12 7 8 9 ...
 $ well_type         : chr [1:27] "sample" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:27] "F05_06_ambient" "F05_01_ambient" "F05_08_ambient" "F05_06_ambient" ...
 $ label             : chr [1:27] "F05_06_ambient-citrate_synthase-30.2-df.0" "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_08_ambient-citrate_synthase-30.6-df.0" "F05_06_ambient-citrate_synthase-30.2-df.0" ...
 $ std_nmol          : num [1:27] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:27] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:27] 0.136 0.135 0.135 0.133 0.131 0.131 0.129 0.128 0.128 0.126 ...
 $ od_last           : num [1:27] 0.323 0.28 0.32 0.319 0.317 0.318 0.314 0.314 0.317 0.311 ...
 $ od_max            : num [1:27] 0.323 0.28 0.32 0.319 0.317 0.318 0.314 0.314 0.317 0.311 ...
 $ net_change        : num [1:27] 0.187 0.145 0.185 0.186 0.186 0.187 0.185 0.186 0.189 0.185 ...
 $ frac_rising       : num [1:27] 1 1 1 1 1 1 1 1 1 1 ...
 $ max_step          : num [1:27] 0.028 0.016 0.032 0.028 0.031 0.029 0.032 0.027 0.029 0.029 ...
 $ typical_step      : num [1:27] 0.02 0.015 0.018 0.0195 0.0185 0.0195 0.018 0.0205 0.02 0.0195 ...
 $ step_excess       : num [1:27] -0.002 -0.003 -0.0015 0.001 -0.002 ...
 $ glitch_at_min     : num [1:27] 20 20 14 10 12 8 12 14 2 2 ...
 $ step_ratio        : num [1:27] 1.4 1.07 1.78 1.44 1.68 ...
 $ flag_decreasing   : logi [1:27] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range   : logi [1:27] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:27] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:27] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:27] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:27] 0 0 0 0 0 0 0 0 0 0 ...
 $ baseline          : chr [1:27] "normal" "normal" "normal" "normal" ...

--- baseline_per_sample: elevated-replicate count per sample ---

tibble [8 × 4] (S3: tbl_df/tbl/data.frame)
 $ sample_id      : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_elevated     : int [1:8] 0 0 0 0 0 0 0 0
 $ n              : int [1:8] 3 3 3 3 3 3 3 3
 $ median_baseline: num [1:8] 0.12 0.119 0.128 0.121 0.119 0.133 0.124 0.131

No sample or background well anywhere on this plate starts above A412 0.35 — sample wells all start within a tight 0.118-0.136 range and the pooled background wells start even lower (0.055-0.064), both well below the 0.35 threshold. No sample has any elevated-baseline replicate, and none is disqualified on this basis.

4.5 Plot starting absorbance

baseline_plot <- ggplot(baseline_check, aes(x = reorder(well, od_first), y = od_first,
                                            fill = well_type)) +
  geom_col() +
  geom_hline(yintercept = assay_params$sample_baseline_od,
             linetype = "dashed", colour = "#cb181d") +
  scale_fill_manual(values = c(sample = "#08519c", background = "#a63603")) +
  labs(title = "Starting A412 (t = 0) for every sample/background well",
       subtitle = paste0("Dashed line = elevated-baseline threshold (A412 = ",
                         assay_params$sample_baseline_od, "). No well on this plate exceeds it."),
       x = "Well", y = "A412 at t = 0", fill = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 90, size = 5, vjust = 0.5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "starting_absorbance.png"), baseline_plot,
       width = 9, height = 6, dpi = 300)

cat("--- baseline_plot: ggplot object structure ---\n\n")
summary(baseline_plot)

baseline_plot

--- baseline_plot: ggplot object structure ---

data: well, plate_row, plate_col, well_type, sample_id, label,
  std_nmol, source, od_first, od_last, od_max, net_change, frac_rising,
  max_step, typical_step, step_excess, glitch_at_min, step_ratio,
  flag_decreasing, flag_over_range, flag_glitch, flag_high_baseline,
  flag_bg_active, n_flags, baseline [27x25]
mapping:  x = ~reorder(well, od_first), y = ~od_first, fill = ~well_type
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

5 GSH STANDARD CURVE

The GSH standard is a stoichiometric endpoint, not an enzymatic reaction: free thiol reduces DTNB immediately on mixing. The standard curve is therefore read from a single timepoint, and t = 0 is used because TNB2- degrades over the run (quantified below).

5.1 Fit the standard curve

Three candidate fits are compared:

  • all wells — every standard replicate, no exclusions
  • concentration means — the conventional fit, on per-concentration means
  • outlier-excluded — replicate-level fit after dropping flagged wells

The outlier-excluded fit is used for activity calculation.

std_curve <- fit_standard_curve(plate_long, assay_params)

standards_t0       <- std_curve$standards_t0
standard_drift     <- std_curve$standard_drift
standard_summary   <- std_curve$standard_summary
standards_flagged  <- std_curve$standards_flagged
fit_all_wells      <- std_curve$fit_all_wells
fit_comparison     <- std_curve$fit_comparison
std_slope          <- std_curve$std_slope
std_intercept      <- std_curve$std_intercept
std_r2             <- std_curve$std_r2
std_nmol_max       <- std_curve$std_nmol_max

cat("Standard wells:", nrow(standards_t0), "at", dplyr::n_distinct(standards_t0$std_nmol),
    "concentrations\n")
cat("Standard drift over 20 min (dA412): median",
    sprintf("%.3f", median(standard_drift$drift)),
    "| range", paste(sprintf("%.3f", range(standard_drift$drift)), collapse = " to "), "\n")
cat("Standards that LOST signal (drift < 0):", sum(standard_drift$drift < 0), "/", nrow(standard_drift), "\n")

cat("\n--- standard_summary: per-concentration mean/SD/CV of t0 absorbance ---\n\n")
str(standard_summary)
kable(standard_summary, digits = c(0, 0, 3, 4, 4, 2, 3, 3, 4),
      caption = "GSH standard curve, per-concentration summary at t = 0")

cat("\nConcentrations exceeding", assay_params$cv_threshold_pct, "% replicate CV:",
    paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
          collapse = ", "), "\n")

cat("\n--- standards_flagged: every standard well with its deviation from the triplicate median ---\n\n")
str(standards_flagged)
kable(standards_flagged %>% filter(is_outlier) %>% select(well, std_nmol, od, triplicate_median, deviation),
      digits = 3, caption = "Standard wells flagged as outliers (|deviation| > 0.15 A412)")

cat("\n--- fit_comparison: slope/intercept/R^2 for all three candidate fits ---\n\n")
str(fit_comparison)
# row.names = FALSE: coef() attaches the predictor's name ("std_nmol") to the
# slope/intercept elements, which data.frame() otherwise promotes to a mangled
# row label (e.g. "std_nmol...1") that breaks pandoc's pipe-table parser.
kable(fit_comparison, digits = c(0, 0, 5, 4, 5), row.names = FALSE,
      caption = "Standard curve fit comparison (outlier-excluded fit used downstream)")
Standard wells: 18 at 6 concentrations
Standard drift over 20 min (dA412): median -0.007 | range -0.030 to 0.003 
Standards that LOST signal (drift < 0): 13 / 18 

--- standard_summary: per-concentration mean/SD/CV of t0 absorbance ---

tibble [6 × 9] (S3: tbl_df/tbl/data.frame)
 $ std_nmol   : num [1:6] 0 8 16 24 32 40
 $ n          : int [1:6] 3 3 3 3 3 3
 $ mean_od    : num [1:6] 0.108 0.376 0.629 0.865 1.118 ...
 $ sd_od      : num [1:6] 0.00153 0.00781 0.00866 0.00351 0.01193 ...
 $ se_od      : num [1:6] 0.000882 0.004509 0.005 0.002028 0.006888 ...
 $ cv_pct     : num [1:6] 1.41 2.077 1.377 0.406 1.067 ...
 $ median_od  : num [1:6] 0.108 0.38 0.634 0.865 1.123 ...
 $ net_od     : num [1:6] 0 0.268 0.521 0.757 1.009 ...
 $ od_per_nmol: num [1:6] NA 0.0335 0.0325 0.0315 0.0315 ...
std_nmol n mean_od sd_od se_od cv_pct median_od net_od od_per_nmol
0 3 0.108 0.0015 0.0009 1.41 0.108 0.000 NA
8 3 0.376 0.0078 0.0045 2.08 0.380 0.268 0.0335
16 3 0.629 0.0087 0.0050 1.38 0.634 0.521 0.0325
24 3 0.865 0.0035 0.0020 0.41 0.865 0.757 0.0315
32 3 1.118 0.0119 0.0069 1.07 1.123 1.009 0.0315
40 3 1.427 0.0074 0.0043 0.52 1.430 1.319 0.0330

GSH standard curve, per-concentration summary at t = 0

Concentrations exceeding 15 % replicate CV:  

--- standards_flagged: every standard well with its deviation from the triplicate median ---

tibble [18 × 7] (S3: tbl_df/tbl/data.frame)
 $ well             : chr [1:18] "D2" "C9" "C4" "D4" ...
 $ std_nmol         : num [1:18] 32 16 8 40 24 40 24 32 0 0 ...
 $ od               : num [1:18] 1.104 0.619 0.367 1.419 0.869 ...
 $ source           : chr [1:18] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ triplicate_median: num [1:18] 1.123 0.634 0.38 1.43 0.865 ...
 $ deviation        : num [1:18] -0.019 -0.015 -0.013 -0.011 0.004 ...
 $ is_outlier       : logi [1:18] FALSE FALSE FALSE FALSE FALSE FALSE ...
well std_nmol od triplicate_median deviation —— ———- —- ——————- ———–

Standard wells flagged as outliers (|deviation| > 0.15 A412)

— fit_comparison: slope/intercept/R^2 for all three candidate fits —

‘data.frame’: 3 obs. of 5 variables: $ fit : chr “all wells” “concentration means” “outlier-excluded” $ n : int 18 6 18 $ slope : num 0.0323 0.0323 0.0323 $ intercept: num 0.107 0.107 0.107 $ r_squared: num 0.998 0.999 0.998

fit n slope intercept r_squared
all wells 18 0.03234 0.1071 0.99831
concentration means 6 0.03234 0.1071 0.99851
outlier-excluded 18 0.03234 0.1071 0.99831

Standard curve fit comparison (outlier-excluded fit used downstream)

No standard well is flagged as an outlier on this plate – every concentration’s triplicate CV is under 2.1% (highest: 8 nmol at 2.1%), and no well deviates from its triplicate median by more than the 0.15 A412 threshold. Because there are no outliers to exclude, the “outlier-excluded” and “all wells” fits are identical: slope = 0.03234 A412/nmol, R2 = 0.9983. Signal decay is more widespread than on prior plates, however: 13 of 15 non-zero-concentration standards lose signal strictly (drift < 0) between t = 0 and t = 20 min (the remaining 2 are unchanged to three decimals), and 15 of 18 are flagged flag_decreasing above once ties at exactly zero net change are included. Because the curve is read at t = 0, this decay affects none of the values used for calibration.

5.2 Plot the standard curve

std_plot <- ggplot(standards_flagged, aes(x = std_nmol, y = od)) +
  geom_point(aes(colour = is_outlier), size = 3) +
  geom_smooth(data = standards_flagged %>% filter(!is_outlier),
              method = "lm", se = TRUE, colour = "#238b45", fill = "#a1d99b") +
  scale_colour_manual(values = c(`TRUE` = "#cb181d", `FALSE` = "#08519c"),
                      labels = c(`TRUE` = "outlier", `FALSE` = "included"),
                      name = "Standard well") +
  labs(title = "GSH standard curve (t = 0), family F05 ambient",
       subtitle = sprintf("Outlier-excluded fit: slope = %.5f A412/nmol, R² = %.4f (n = %d)",
                          std_slope, std_r2, nrow(standards_flagged %>% filter(!is_outlier))),
       x = "GSH (nmol/well)", y = "A412 at t = 0") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "gsh_standard_curve.png"), std_plot,
       width = 9, height = 6, dpi = 300)

cat("--- std_plot: ggplot object structure ---\n\n")
summary(std_plot)

std_plot

--- std_plot: ggplot object structure ---

data: well, std_nmol, od, source, triplicate_median, deviation,
  is_outlier [18x7]
mapping:  x = ~std_nmol, y = ~od
scales:   colour 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~is_outlier 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

geom_smooth: na.rm = FALSE, orientation = NA, se = TRUE
stat_smooth: na.rm = FALSE, orientation = NA, se = TRUE, method = lm
position_identity

The top standard (40 nmol) stays comfortably within the photometric linear range here – the highest reading is A412 1.43, below the 1.5 ceiling – so the calibration does not need to be restricted to a reduced concentration range.

6 RATE EXTRACTION

Rates are extracted with a sliding-window approach: for each well, the 5-point (8-minute) window of steepest positive slope is taken as the rate, with its R2 retained for the usability check.

well_rates <- compute_well_rates(plate_long, well_diagnostics, assay_params)

cat("--- well_rates: extracted rate window per well ---\n\n")
str(well_rates)
--- well_rates: extracted rate window per well ---

tibble [48 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:48] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:48] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:48] "F05_01_ambient" "F05_04_ambient" "F05_04_ambient" "F05_04_ambient" ...
 $ std_nmol              : num [1:48] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:48] 4 2 0 0 2 2 4 2 2 0 ...
 $ t_end                 : num [1:48] 12 10 8 8 10 10 12 10 10 8 ...
 $ slope_mOD_min         : num [1:48] 7.75 9.9 10.25 10.4 7.85 ...
 $ r2                    : num [1:48] 1 1 1 1 1 ...
 $ max_abs_slope_mOD_min : num [1:48] 7.75 9.9 10.25 10.4 7.85 ...
 $ abs_window_is_negative: logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:48] 0.145 0.172 0.177 0.179 0.148 0.138 0.155 0.161 0.152 0.186 ...
 $ frac_rising           : num [1:48] 1 1 1 1 1 1 1 1 1 1 ...
 $ od_max                : num [1:48] 0.28 0.293 0.297 0.301 0.268 0.257 0.274 0.285 0.271 0.314 ...
 $ flag_decreasing       : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:48] 20 20 20 20 2 2 20 20 2 14 ...
 $ flag_bg_active        : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:48] 0 0 0 0 0 0 0 0 0 0 ...
 $ glitch_in_window      : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:48] TRUE TRUE TRUE TRUE TRUE TRUE ...

6.1 Rate tables by well type

sample_rates <- well_rates %>% filter(well_type == "sample")

kable(sample_rates %>% select(well, sample_id, t_start, t_end, slope_mOD_min, r2, rate_usable),
      digits = c(0, 0, 0, 0, 3, 4, 0),
      caption = "Extracted rate window, sample wells")

cat("\nSample wells with a usable rate:", sum(sample_rates$rate_usable), "/", nrow(sample_rates), "\n")
cat("Sample rate range (mA412/min):",
    paste(sprintf("%.2f", range(sample_rates$slope_mOD_min)), collapse = " - "), "\n")
cat("Median replicate-level R^2:", sprintf("%.4f", median(sample_rates$r2)), "\n")

cat("\n--- sample_rates: rate-window results restricted to sample wells ---\n\n")
str(sample_rates)
well sample_id t_start t_end slope_mOD_min r2 rate_usable
A1 F05_01_ambient 4 12 7.75 0.9999 TRUE
A10 F05_04_ambient 2 10 9.90 0.9999 TRUE
A11 F05_04_ambient 0 8 10.25 0.9999 TRUE
A12 F05_04_ambient 0 8 10.40 0.9999 TRUE
A2 F05_01_ambient 2 10 7.85 0.9999 TRUE
A3 F05_01_ambient 2 10 7.20 0.9998 TRUE
A4 F05_02_ambient 4 12 8.35 0.9999 TRUE
A5 F05_02_ambient 2 10 8.90 0.9999 TRUE
A6 F05_02_ambient 2 10 8.25 0.9999 TRUE
A7 F05_03_ambient 0 8 13.15 0.9997 TRUE
A8 F05_03_ambient 0 8 13.70 0.9997 TRUE
A9 F05_03_ambient 0 8 13.60 0.9994 TRUE
B1 F05_05_ambient 6 14 5.65 0.9998 TRUE
B10 F05_08_ambient 0 8 14.80 0.9984 TRUE
B11 F05_08_ambient 0 8 14.60 0.9982 TRUE
B12 F05_08_ambient 0 8 14.80 0.9984 TRUE
B2 F05_05_ambient 10 18 5.70 0.9997 TRUE
B3 F05_05_ambient 10 18 5.50 1.0000 TRUE
B4 F05_06_ambient 0 8 13.40 0.9996 TRUE
B5 F05_06_ambient 0 8 13.55 0.9995 TRUE
B6 F05_06_ambient 0 8 13.90 0.9994 TRUE
B7 F05_07_ambient 0 8 11.15 0.9999 TRUE
B8 F05_07_ambient 0 8 11.45 0.9995 TRUE
B9 F05_07_ambient 0 8 11.30 0.9997 TRUE

Extracted rate window, sample wells

Sample wells with a usable rate: 24 / 24 
Sample rate range (mA412/min): 5.50 - 14.80 
Median replicate-level R^2: 0.9997 

--- sample_rates: rate-window results restricted to sample wells ---

tibble [24 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:24] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:24] "F05_01_ambient" "F05_04_ambient" "F05_04_ambient" "F05_04_ambient" ...
 $ std_nmol              : num [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:24] 4 2 0 0 2 2 4 2 2 0 ...
 $ t_end                 : num [1:24] 12 10 8 8 10 10 12 10 10 8 ...
 $ slope_mOD_min         : num [1:24] 7.75 9.9 10.25 10.4 7.85 ...
 $ r2                    : num [1:24] 1 1 1 1 1 ...
 $ max_abs_slope_mOD_min : num [1:24] 7.75 9.9 10.25 10.4 7.85 ...
 $ abs_window_is_negative: logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:24] 0.145 0.172 0.177 0.179 0.148 0.138 0.155 0.161 0.152 0.186 ...
 $ frac_rising           : num [1:24] 1 1 1 1 1 1 1 1 1 1 ...
 $ od_max                : num [1:24] 0.28 0.293 0.297 0.301 0.268 0.257 0.274 0.285 0.271 0.314 ...
 $ flag_decreasing       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:24] 20 20 20 20 2 2 20 20 2 14 ...
 $ flag_bg_active        : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:24] 0 0 0 0 0 0 0 0 0 0 ...
 $ glitch_in_window      : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:24] TRUE TRUE TRUE TRUE TRUE TRUE ...

Every one of the 24 sample wells clears the R2 >= 0.80 and positive-net-change usability bar; every replicate-level R2 exceeds 0.998 (median 0.9997). No well needed to be excluded from rate extraction on this plate – a direct consequence of the clean, monotonically-rising traces noted above. Sample rates range 5.50-14.80 mA412/min, a considerably wider spread than the raw rates seen on this same family’s 36 °C plate, reflecting real inter-individual variation among these ambient-temperature homogenates (see CITRATE SYNTHASE ACTIVITY below).

6.2 Positive control

pos_control <- compute_positive_control(well_rates, std_curve$std_slope, assay_params)

kable(pos_control %>% select(well, t_start, t_end, slope_mOD_min, r2, net_change, activity_mU_uL),
      digits = c(0, 0, 0, 3, 4, 3, 4),
      caption = "Positive control replicates")

cat("Positive control mean rate:", sprintf("%.3f", mean(pos_control$slope_mOD_min)), "mA412/min\n")
cat("Positive control CV:", sprintf("%.1f",
    100 * sd(pos_control$slope_mOD_min) / mean(pos_control$slope_mOD_min)), "%\n")
cat("All replicates rising and linear (R^2 > 0.99):",
    all(pos_control$net_change > 0 & pos_control$r2 > 0.99), "\n")

cat("\n--- pos_control: positive control rate and R^2 per replicate ---\n\n")
str(pos_control)
well t_start t_end slope_mOD_min r2 net_change activity_mU_uL
D7 12 20 2.00 1.0000 0.036 0.0309
D8 10 18 2.95 0.9968 0.057 0.0456
D9 12 20 3.15 0.9992 0.059 0.0487

Positive control replicates

Positive control mean rate: 2.700 mA412/min
Positive control CV: 22.8 %
All replicates rising and linear (R^2 > 0.99): TRUE 

--- pos_control: positive control rate and R^2 per replicate ---

tibble [3 × 22] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:3] "D7" "D8" "D9"
 $ well_type             : chr [1:3] "positive_control" "positive_control" "positive_control"
 $ sample_id             : chr [1:3] NA NA NA
 $ std_nmol              : num [1:3] NA NA NA
 $ t_start               : num [1:3] 12 10 12
 $ t_end                 : num [1:3] 20 18 20
 $ slope_mOD_min         : num [1:3] 2 2.95 3.15
 $ r2                    : num [1:3] 1 0.997 0.999
 $ max_abs_slope_mOD_min : num [1:3] 2 2.95 3.15
 $ abs_window_is_negative: logi [1:3] FALSE FALSE FALSE
 $ net_change            : num [1:3] 0.036 0.057 0.059
 $ frac_rising           : num [1:3] 1 1 1
 $ od_max                : num [1:3] 0.202 0.317 0.34
 $ flag_decreasing       : logi [1:3] FALSE FALSE FALSE
 $ flag_over_range       : logi [1:3] FALSE FALSE FALSE
 $ flag_glitch           : logi [1:3] FALSE FALSE FALSE
 $ glitch_at_min         : num [1:3] 2 2 2
 $ flag_bg_active        : logi [1:3] FALSE FALSE FALSE
 $ n_flags               : int [1:3] 0 0 0
 $ glitch_in_window      : logi [1:3] FALSE FALSE FALSE
 $ rate_usable           : logi [1:3] TRUE TRUE TRUE
 $ activity_mU_uL        : num [1:3] 0.0309 0.0456 0.0487

The positive control rose linearly in all three replicates (R2 0.997-1.000), confirming that the core reaction chemistry and reader were functioning. Its replicate-to-replicate precision is notably worse than on prior plates, though: D7 rose at only 2.00 mA412/min versus 2.95 and 3.15 mA412/min for D8/D9, a 22.8% CV across the triplicate (well above the 15% technical-CV threshold applied to samples). Because the positive control is a validity check on assay chemistry, not an input to any sample’s activity calculation, this does not affect the results table below – but it is flagged again in the SUMMARY as a caution for future plates.

7 BACKGROUND CORRECTION

Unlike the 2026-08-11 and 2026-08-14 plates, which paired each sample with its own triplicate background control, this plate uses a single pooled background control (note 3): the background rate is estimated once, from the flat (well-behaved) replicates of the BG-citrate_synthase triplicate, and the same corrected rate is subtracted from every sample.

background_correction <- compute_background_correction(well_rates)

cat("Pooled background rate (flat wells only):\n")
print(background_correction, row.names = FALSE, digits = 3)

cat("\nBackground rate used for correction:",
    sprintf("%.3f", background_correction$bg_rate_flat), "mA412/min (from",
    background_correction$n_bg_flat, "of", background_correction$n_bg_total, "replicates)\n")
cat("Median sample rate:",
    sprintf("%.2f", median(sample_rates$slope_mOD_min)), "mA412/min\n")
cat("Background as % of median sample signal:",
    sprintf("%.1f", 100 * background_correction$bg_rate_flat /
                    median(sample_rates$slope_mOD_min)), "%\n")

cat("\n--- background_correction: pooled background rate estimate ---\n\n")
str(background_correction)
Pooled background rate (flat wells only):
 n_bg_total n_bg_flat bg_rate_flat bg_rate_all
          3         3        0.233       0.233

Background rate used for correction: 0.233 mA412/min (from 3 of 3 replicates)
Median sample rate: 10.78 mA412/min
Background as % of median sample signal: 2.2 %

--- background_correction: pooled background rate estimate ---

'data.frame':   1 obs. of  4 variables:
 $ n_bg_total  : int 3
 $ n_bg_flat   : int 3
 $ bg_rate_flat: num 0.233
 $ bg_rate_all : num 0.233

All three background replicates are flat and are used in the estimate (0.233 mA412/min), a much smaller fraction of the sample signal (~2.2%) than the paired per-sample backgrounds on the 36 °C plate (~7%) – a consequence of the pooled background containing only 2 µL of a 1-in-8 pooled homogenate per well, rather than the full undiluted 2 µL of a single sample’s own homogenate.

7.1 Background significance test

The flat/active threshold classification above is backed by an explicit significance test: a Welch two-sample t-test of well-level rate (background vs. sample), plus an ANOVA on the OD trajectory over time. This mirrors the one-time background ANOVA validation done by Cattau et al. (2023) on the legacy version of this assay.

bg_sample_rates <- well_rates %>%
  filter(well_type %in% c("sample", "background"))

bg_significance_test <- t.test(slope_mOD_min ~ well_type, data = bg_sample_rates)

cat("--- Welch two-sample t-test: well-level rate (background vs. sample) ---\n\n")
print(bg_significance_test)

bg_trend_anova <- anova(lm(od ~ well_type * time_min,
                            data = plate_long %>% filter(well_type %in% c("sample", "background"))))

cat("\n--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---\n\n")
print(bg_trend_anova)

cat("\nInterpretation: mean background-well rate is",
    sprintf("%.2f", unname(bg_significance_test$estimate["mean in group background"])),
    "mA412/min vs.", sprintf("%.2f", unname(bg_significance_test$estimate["mean in group sample"])),
    "mA412/min for sample wells (p =", format.pval(bg_significance_test$p.value, digits = 3),
    "). Background wells are statistically distinct from, and far slower\n",
    "than, reaction wells on this plate -- consistent with the pooled\n",
    "flat/active classification above.\n")

cat("\n--- bg_significance_test: Welch t-test result object ---\n\n")
str(bg_significance_test)

cat("\n--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---\n\n")
str(bg_trend_anova)
--- Welch two-sample t-test: well-level rate (background vs. sample) ---


    Welch Two Sample t-test

data:  slope_mOD_min by well_type
t = -16.416, df = 23.032, p-value = 3.317e-14
alternative hypothesis: true difference in means between group background and group sample is not equal to 0
95 percent confidence interval:
 -11.705749  -9.085918
sample estimates:
mean in group background     mean in group sample 
               0.2333333               10.6291667 


--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---

Analysis of Variance Table

Response: od
                    Df  Sum Sq Mean Sq F value    Pr(>F)    
well_type            1 0.72937 0.72937 1026.15 < 2.2e-16 ***
time_min             1 0.66510 0.66510  935.74 < 2.2e-16 ***
well_type:time_min   1 0.07971 0.07971  112.14 < 2.2e-16 ***
Residuals          293 0.20826 0.00071                      
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Interpretation: mean background-well rate is 0.23 mA412/min vs. 10.63 mA412/min for sample wells (p = 3.32e-14 ). Background wells are statistically distinct from, and far slower
 than, reaction wells on this plate -- consistent with the pooled
 flat/active classification above.

--- bg_significance_test: Welch t-test result object ---

List of 10
 $ statistic  : Named num -16.4
  ..- attr(*, "names")= chr "t"
 $ parameter  : Named num 23
  ..- attr(*, "names")= chr "df"
 $ p.value    : num 3.32e-14
 $ conf.int   : num [1:2] -11.71 -9.09
  ..- attr(*, "conf.level")= num 0.95
 $ estimate   : Named num [1:2] 0.233 10.629
  ..- attr(*, "names")= chr [1:2] "mean in group background" "mean in group sample"
 $ null.value : Named num 0
  ..- attr(*, "names")= chr "difference in means between group background and group sample"
 $ stderr     : num 0.633
 $ alternative: chr "two.sided"
 $ method     : chr "Welch Two Sample t-test"
 $ data.name  : chr "slope_mOD_min by well_type"
 - attr(*, "class")= chr "htest"

--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---

Classes 'anova' and 'data.frame':   4 obs. of  5 variables:
 $ Df     : int  1 1 1 293
 $ Sum Sq : num  0.7294 0.6651 0.0797 0.2083
 $ Mean Sq: num  0.729366 0.665104 0.079706 0.000711
 $ F value: num  1026 936 112 NA
 $ Pr(>F) : num  9.90e-98 3.29e-93 2.12e-22 NA
 - attr(*, "heading")= chr [1:2] "Analysis of Variance Table\n" "Response: od"

8 TECHNICAL REPLICATE PRECISION

Coefficient of variation across the three technical replicates of each sample, computed on the extracted rate. Both all and usable replicates only are reported; here they are identical for every sample because no replicate was excluded.

cv_summary <- compute_replicate_cv(well_rates, protein_by_sample$sample_id, assay_params)

kable(cv_summary %>% select(sample_id, n_all, mean_all, sd_all, cv_all,
                            n_usable, mean_usable, sd_usable, cv_usable,
                            excluded_wells),
      digits = c(0, 0, 2, 2, 1, 0, 2, 2, 1, 0),
      col.names = c("Sample", "n", "Mean rate", "SD", "CV (%)",
                    "n usable", "Mean rate", "SD", "CV (%)", "Excluded wells"),
      caption = paste0("Technical replicate CV of the extracted rate (mA412/min), ",
                       "all replicates vs. usable replicates only"))

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on ALL replicates:", sum(cv_summary$fails_cv_all), "/", nrow(cv_summary), "\n")
print(cv_summary %>% filter(fails_cv_all) %>%
        select(sample_id, n_all, mean_all, sd_all, cv_all) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on USABLE replicates:", sum(cv_summary$fails_cv_usable), "\n")
print(cv_summary %>% filter(fails_cv_usable) %>%
        select(sample_id, n_usable, mean_usable, sd_usable, cv_usable) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

write.csv(cv_summary, file.path(output_dir, "technical_replicate_cv.csv"), row.names = FALSE)

cat("\n--- cv_summary: per-sample technical replicate CV, both variants ---\n\n")
str(cv_summary)
Sample n Mean rate SD CV (%) n usable Mean rate SD CV (%) Excluded wells
F05_01_ambient 3 7.60 0.35 4.6 3 7.60 0.35 4.6 -
F05_02_ambient 3 8.50 0.35 4.1 3 8.50 0.35 4.1 -
F05_03_ambient 3 13.48 0.29 2.2 3 13.48 0.29 2.2 -
F05_04_ambient 3 10.18 0.26 2.5 3 10.18 0.26 2.5 -
F05_05_ambient 3 5.62 0.10 1.9 3 5.62 0.10 1.9 -
F05_06_ambient 3 13.62 0.26 1.9 3 13.62 0.26 1.9 -
F05_07_ambient 3 11.30 0.15 1.3 3 11.30 0.15 1.3 -
F05_08_ambient 3 14.73 0.12 0.8 3 14.73 0.12 0.8 -

Technical replicate CV of the extracted rate (mA412/min), all replicates vs. usable replicates only

Samples exceeding CV 15% on ALL replicates: 0 / 8 
[1] sample_id n_all     mean_all  sd_all    cv_all   
<0 rows> (or 0-length row.names)

Samples exceeding CV 15% on USABLE replicates: 0 
[1] sample_id   n_usable    mean_usable sd_usable   cv_usable  
<0 rows> (or 0-length row.names)

--- cv_summary: per-sample technical replicate CV, both variants ---

'data.frame':   8 obs. of  12 variables:
 $ sample_id      : chr  "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_all          : int  3 3 3 3 3 3 3 3
 $ mean_all       : num  7.6 8.5 13.48 10.18 5.62 ...
 $ sd_all         : num  0.35 0.35 0.293 0.257 0.104 ...
 $ cv_all         : num  4.61 4.12 2.17 2.52 1.85 ...
 $ n_usable       : int  3 3 3 3 3 3 3 3
 $ mean_usable    : num  7.6 8.5 13.48 10.18 5.62 ...
 $ sd_usable      : num  0.35 0.35 0.293 0.257 0.104 ...
 $ cv_usable      : num  4.61 4.12 2.17 2.52 1.85 ...
 $ excluded_wells : Named chr  "-" "-" "-" "-" ...
  ..- attr(*, "names")= chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ fails_cv_all   : logi  FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ fails_cv_usable: logi  FALSE FALSE FALSE FALSE FALSE FALSE ...

No sample exceeds the 15% CV threshold on this plate. The highest is F05_01_ambient at 4.6%; the lowest is F05_08_ambient at 0.8%, and every sample falls in the 0.8-4.6% range – tighter than the positive control’s own 22.8% CV noted above.

8.1 Plot replicate spread

cv_plot_data <- sample_rates %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id")

cv_plot <- ggplot(cv_plot_data, aes(x = sample_id, y = slope_mOD_min)) +
  geom_point(aes(colour = rate_usable, shape = rate_usable),
             size = 3, position = position_jitter(width = 0.12, height = 0)) +
  stat_summary(fun = mean, geom = "crossbar", width = 0.5,
               colour = "grey30", linewidth = 0.3) +
  geom_text(data = cv_summary, aes(x = sample_id, y = Inf,
                                   label = sprintf("CV %.0f%%", cv_all)),
            vjust = 1.6, size = 3,
            colour = ifelse(cv_summary$fails_cv_all, "#cb181d", "grey30")) +
  scale_colour_manual(values = c(`TRUE` = "#08519c", `FALSE` = "#cb181d"),
                      labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                     labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  labs(title = "Technical replicate spread in extracted rate",
       subtitle = paste0("Crossbar = mean of all three replicates; CV computed on all three. ",
                         "Threshold ", assay_params$cv_threshold_pct, "%."),
       x = "Sample", y = "Rate (mA412/min)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 30, hjust = 1))

ggsave(file.path(output_dir, "technical_replicate_cv.png"), cv_plot,
       width = 9, height = 6, dpi = 300)

cat("--- cv_plot: ggplot object structure ---\n\n")
summary(cv_plot)

cv_plot

--- cv_plot: ggplot object structure ---

data: well, well_type, sample_id, std_nmol, t_start, t_end,
  slope_mOD_min, r2, max_abs_slope_mOD_min, abs_window_is_negative,
  net_change, frac_rising, od_max, flag_decreasing, flag_over_range,
  flag_glitch, glitch_at_min, flag_bg_active, n_flags,
  glitch_in_window, rate_usable, cv_all [24x22]
mapping:  x = ~sample_id, y = ~slope_mOD_min
scales:   colour, shape 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~rate_usable, shape = ~rate_usable 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_jitter 

geom_crossbar: na.rm = FALSE, orientation = NA, width = 0.5
stat_summary: fun.data = NULL, fun = function (x, ...) 
UseMethod("mean"), fun.max = NULL, fun.min = NULL, fun.args = list(), na.rm = FALSE, orientation = NA
position_identity 

mapping: x = ~sample_id, y = Inf, label = ~sprintf("CV %.0f%%", cv_all) 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge

9 CITRATE SYNTHASE ACTIVITY

9.1 Calculation

Following Abcam Section 10.3:

Sample CS activity = B / (dT x V) x D nmol/min/µL (= mU/µL)

  1. Rate (A412/min) = max-increasing sliding-window slope / 1000
  2. Background-corrected rate = sample rate - the single pooled background rate (note 3), applied identically to every sample
  3. B/dT (nmol/min) = background-corrected rate / standard curve slope (0.03234 A412 per nmol, from the outlier-excluded fit)
  4. Activity (mU/µL) = (B/dT) / V x D, with V = 2 µL and D = 1
  5. Protein-normalized (mU/mg protein) = activity x homogenate_volume_uL (350 µL, measured) / total protein (mg)
cs_activity <- calculate_cs_activity(well_rates, protein_by_sample, background_correction,
                                      std_slope, std_nmol_max, assay_params, plate_long)

cat("Samples with a computed activity:", nrow(cs_activity), "/", nrow(protein_by_sample), "\n")
cat("Activity range (mU/uL):",
    paste(sprintf("%.4f", range(cs_activity$activity_mU_per_uL)), collapse = " - "), "\n")
cat("Activity range (mU/mg protein):",
    paste(sprintf("%.3f", range(cs_activity$activity_mU_per_mg_protein)), collapse = " - "), "\n")

cat("\n--- cs_activity: per-sample citrate synthase activity ---\n\n")
str(cs_activity)
Samples with a computed activity: 8 / 8 
Activity range (mU/uL): 0.0832 - 0.2242 
Activity range (mU/mg protein): 82.906 - 136.192 

--- cs_activity: per-sample citrate synthase activity ---

tibble [8 × 19] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_reps_used               : int [1:8] 3 3 3 3 3 3 3 3
 $ mean_rate_mOD_min         : num [1:8] 7.6 8.5 13.48 10.18 5.62 ...
 $ sd_rate                   : num [1:8] 0.35 0.35 0.293 0.257 0.104 ...
 $ cv_rate                   : num [1:8] 4.61 4.12 2.17 2.52 1.85 ...
 $ family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ individual                : chr [1:8] "01" "02" "03" "04" ...
 $ temperature               : chr [1:8] "ambient" "ambient" "ambient" "ambient" ...
 $ conc_ug_mL                : num [1:8] 876 1039 1900 1129 712 ...
 $ total_protein_mg          : num [1:8] 0.307 0.364 0.665 0.395 0.249 ...
 $ bg_rate_mOD_min           : num [1:8] 0.233 0.233 0.233 0.233 0.233 ...
 $ corrected_mOD_min         : num [1:8] 7.37 8.27 13.25 9.95 5.38 ...
 $ rate_OD_min               : num [1:8] 0.00737 0.00827 0.01325 0.00995 0.00538 ...
 $ nmol_per_min              : num [1:8] 0.228 0.256 0.41 0.308 0.166 ...
 $ activity_mU_per_uL        : num [1:8] 0.1139 0.1278 0.2048 0.1538 0.0832 ...
 $ total_mU_in_homogenate    : num [1:8] 39.9 44.7 71.7 53.8 29.1 ...
 $ activity_mU_per_mg_protein: num [1:8] 130 123 108 136 117 ...
 $ nmol_in_window            : num [1:8] 4.56 5.11 8.19 6.15 3.33 ...
 $ within_std_range          : logi [1:8] TRUE TRUE TRUE TRUE TRUE TRUE ...

9.2 Results table

results_table <- build_results_table(cs_activity, cv_summary, baseline_per_sample, assay_params)

kable(results_table,
      caption = paste0("Citrate synthase activity, family F05 ambient (mU/mg protein uses ",
                       "measured protein concentration x the ", assay_params$homogenate_volume_uL,
                       " uL homogenization volume)"))

cat("\nInterpretation summary:\n")
print(results_table %>% count(Interpretation, name = "n_samples") %>%
        as.data.frame(), row.names = FALSE)

write.csv(results_table, file.path(output_dir, "citrate_synthase_activity_results.csv"),
          row.names = FALSE)

cat("\n--- results_table: formatted per-sample results written to CSV ---\n\n")
str(results_table)
Sample Family Individual Temperature Protein conc (ug/mL) Total protein (mg) Reps used CV all reps (%) CV used reps (%) Rate (mA412/min) BG rate (mA412/min) Corrected rate (mA412/min) Activity (mU/uL) Activity (mU/mg protein) CV flag Elevated baseline reps Interpretation
F05_01_ambient F05 01 ambient 876.1 0.307 3/3 4.6 4.6 7.60 0.233 7.37 0.1139 129.985 pass 0/3 usable
F05_02_ambient F05 02 ambient 1038.9 0.364 3/3 4.1 4.1 8.50 0.233 8.27 0.1278 123.008 pass 0/3 usable
F05_03_ambient F05 03 ambient 1900.4 0.665 3/3 2.2 2.2 13.48 0.233 13.25 0.2048 107.782 pass 0/3 usable
F05_04_ambient F05 04 ambient 1129.4 0.395 3/3 2.5 2.5 10.18 0.233 9.95 0.1538 136.192 pass 0/3 usable
F05_05_ambient F05 05 ambient 711.7 0.249 3/3 1.9 1.9 5.62 0.233 5.38 0.0832 116.931 pass 0/3 usable
F05_06_ambient F05 06 ambient 2232.4 0.781 3/3 1.9 1.9 13.62 0.233 13.38 0.2069 92.676 pass 0/3 usable
F05_07_ambient F05 07 ambient 1687.8 0.591 3/3 1.3 1.3 11.30 0.233 11.07 0.1711 101.361 pass 0/3 usable
F05_08_ambient F05 08 ambient 2703.7 0.946 3/3 0.8 0.8 14.73 0.233 14.50 0.2242 82.906 pass 0/3 usable

Citrate synthase activity, family F05 ambient (mU/mg protein uses measured protein concentration x the 350 uL homogenization volume)

Interpretation summary:
 Interpretation n_samples
         usable         8

--- results_table: formatted per-sample results written to CSV ---

tibble [8 × 17] (S3: tbl_df/tbl/data.frame)
 $ Sample                    : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ Family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ Individual                : chr [1:8] "01" "02" "03" "04" ...
 $ Temperature               : chr [1:8] "ambient" "ambient" "ambient" "ambient" ...
 $ Protein conc (ug/mL)      : num [1:8] 876 1039 1900 1129 712 ...
 $ Total protein (mg)        : num [1:8] 0.307 0.364 0.665 0.395 0.249 0.781 0.591 0.946
 $ Reps used                 : chr [1:8] "3/3" "3/3" "3/3" "3/3" ...
 $ CV all reps (%)           : num [1:8] 4.6 4.1 2.2 2.5 1.9 1.9 1.3 0.8
 $ CV used reps (%)          : num [1:8] 4.6 4.1 2.2 2.5 1.9 1.9 1.3 0.8
 $ Rate (mA412/min)          : num [1:8] 7.6 8.5 13.48 10.18 5.62 ...
 $ BG rate (mA412/min)       : num [1:8] 0.233 0.233 0.233 0.233 0.233 0.233 0.233 0.233
 $ Corrected rate (mA412/min): num [1:8] 7.37 8.27 13.25 9.95 5.38 ...
 $ Activity (mU/uL)          : num [1:8] 0.1139 0.1278 0.2048 0.1538 0.0832 ...
 $ Activity (mU/mg protein)  : num [1:8] 130 123 108 136 117 ...
 $ CV flag                   : chr [1:8] "pass" "pass" "pass" "pass" ...
 $ Elevated baseline reps    : chr [1:8] "0/3" "0/3" "0/3" "0/3" ...
 $ Interpretation            : chr [1:8] "usable" "usable" "usable" "usable" ...

All eight samples are usable on this plate – no CV flag, no elevated baseline replicate.

9.3 Plot activity

# Three-level quality scheme, kept for consistency with other plate
# documents even though no sample here falls in the flagged categories.
qual_levels <- c("clean", paste0("CV > ", assay_params$cv_threshold_pct, "%"),
                 "all reps baseline-compromised")
qual_cols   <- setNames(c("#08519c", "#cb181d", "grey60"), qual_levels)

activity_plot_data <- cs_activity %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id") %>%
  left_join(baseline_per_sample %>% select(sample_id, n_elevated, n), by = "sample_id") %>%
  mutate(cv_flag = factor(case_when(
                            n_elevated == n                              ~ qual_levels[3],
                            cv_all > assay_params$cv_threshold_pct       ~ qual_levels[2],
                            TRUE                                        ~ qual_levels[1]),
                          levels = qual_levels),
         label_n = paste0(n_reps_used, "/3 reps",
                          ifelse(n_elevated == n, " - DISQUALIFIED", "")))

activity_plot <- ggplot(activity_plot_data,
                        aes(x = reorder(sample_id, activity_mU_per_uL),
                            y = activity_mU_per_uL, fill = cv_flag)) +
  geom_col(width = 0.65) +
  geom_text(aes(label = label_n), hjust = -0.15, size = 3, colour = "grey25") +
  scale_fill_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  coord_flip(clip = "off") +
  expand_limits(y = max(activity_plot_data$activity_mU_per_uL) * 1.25) +
  labs(title = "Citrate synthase activity, M. gigas ctenidia, family F05 ambient",
       subtitle = "Background-corrected, mean of all three technical replicates (all usable on this plate).",
       x = NULL, y = "CS activity (mU/uL homogenate)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "citrate_synthase_activity.png"), activity_plot,
       width = 10, height = 6, dpi = 300)

cat("--- activity_plot: ggplot object structure ---\n\n")
summary(activity_plot)

activity_plot

--- activity_plot: ggplot object structure ---

data: sample_id, n_reps_used, mean_rate_mOD_min, sd_rate, cv_rate,
  family, individual, temperature, conc_ug_mL, total_protein_mg,
  bg_rate_mOD_min, corrected_mOD_min, rate_OD_min, nmol_per_min,
  activity_mU_per_uL, total_mU_in_homogenate,
  activity_mU_per_mg_protein, nmol_in_window, within_std_range, cv_all,
  n_elevated, n, cv_flag, label_n [8x24]
mapping:  x = ~reorder(sample_id, activity_mU_per_uL), y = ~activity_mU_per_uL, fill = ~cv_flag
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: label = ~label_n 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge 

mapping: y = ~y 
geom_blank: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

Raw activity (mU/µL) spans a real 2.7-fold range across the eight individuals – 0.0832 (F05_05_ambient) to 0.2242 (F05_08_ambient) – against a technical CV of under 5% for any single sample, so this spread reflects genuine plate-to-plate biological/extraction variation rather than measurement noise.

9.4 Protein-normalized activity

# All eight samples are clean on this plate, so the trend fit uses all of them.
norm_clean <- activity_plot_data %>% filter(cv_flag == qual_levels[1])

norm_plot <- ggplot(activity_plot_data,
                    aes(x = total_protein_mg, y = activity_mU_per_mg_protein)) +
  geom_smooth(data = norm_clean, method = "lm", se = TRUE,
              colour = "grey35", fill = "grey85", linewidth = 0.6) +
  geom_point(aes(colour = cv_flag), size = 3.5) +
  geom_text(aes(label = individual), vjust = -1, size = 3, colour = "grey25") +
  scale_colour_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  labs(title = "Protein-normalized CS activity vs. total extracted protein",
       subtitle = paste0("Point labels = individual ID. Total protein = measured protein ",
                         "concentration x the same\n", assay_params$homogenate_volume_uL,
                         " uL homogenization volume used for every sample. Grey line = fit to ",
                         "all 8\nsamples (all clean on this plate); a positive raw-rate/protein ",
                         "relationship would suggest\nnormalization is correcting for real ",
                         "extraction-efficiency differences rather than introducing an artifact."),
       x = "Total extracted protein (mg)", y = "CS activity (mU/mg protein)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

protein_trend <- cor.test(activity_plot_data$total_protein_mg, activity_plot_data$activity_mU_per_mg_protein)
protein_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_mg_protein)
uL_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_uL)

ggsave(file.path(output_dir, "citrate_synthase_activity_normalized.png"), norm_plot,
       width = 10, height = 6, dpi = 300)

cat("Correlation of mU/mg protein with total extracted protein\n")
cat(sprintf("  all %d samples      : r = %+.3f, p = %.4f\n",
            nrow(activity_plot_data), protein_trend$estimate, protein_trend$p.value))
cat(sprintf("  clean samples only  : r = %+.3f, p = %.4f  (n = %d)\n",
            protein_trend_clean$estimate, protein_trend_clean$p.value, nrow(norm_clean)))
cat(sprintf("  clean-sample mU/mg protein range: %.3f to %.3f (%.2f-fold)\n",
            min(norm_clean$activity_mU_per_mg_protein), max(norm_clean$activity_mU_per_mg_protein),
            max(norm_clean$activity_mU_per_mg_protein) / min(norm_clean$activity_mU_per_mg_protein)))
cat(sprintf("  clean-sample mU/uL range: %.4f to %.4f (%.2f-fold)\n",
            min(norm_clean$activity_mU_per_uL), max(norm_clean$activity_mU_per_uL),
            max(norm_clean$activity_mU_per_uL) / min(norm_clean$activity_mU_per_uL)))
cat(sprintf("\nArtifact check: r(total protein, mU/uL raw rate) = %+.3f, p = %.4f  (n = %d)\n",
            uL_trend_clean$estimate, uL_trend_clean$p.value, nrow(norm_clean)))
cat("If mU/uL itself trends with total protein while mU/mg protein trends oppositely,\n",
    "the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate\n",
    "-- but a STRONG raw-rate/protein relationship (as seen here) instead argues that\n",
    "protein normalization is correcting for genuine extraction-efficiency differences.\n", sep = "")

cat("\n--- protein_trend_clean: correlation test on clean samples ---\n\n")
str(protein_trend_clean, max.level = 1, give.attr = FALSE)

cat("\n--- norm_clean: clean samples used for the trend fit ---\n\n")
str(norm_clean)

norm_plot

Correlation of mU/mg protein with total extracted protein
  all 8 samples      : r = -0.888, p = 0.0032
  clean samples only  : r = -0.888, p = 0.0032  (n = 8)
  clean-sample mU/mg protein range: 82.906 to 136.192 (1.64-fold)
  clean-sample mU/uL range: 0.0832 to 0.2242 (2.69-fold)

Artifact check: r(total protein, mU/uL raw rate) = +0.956, p = 0.0002  (n = 8)
If mU/uL itself trends with total protein while mU/mg protein trends oppositely,
the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate
-- but a STRONG raw-rate/protein relationship (as seen here) instead argues that
protein normalization is correcting for genuine extraction-efficiency differences.

--- protein_trend_clean: correlation test on clean samples ---

List of 9
 $ statistic  : Named num -4.74
 $ parameter  : Named int 6
 $ p.value    : num 0.00318
 $ estimate   : Named num -0.888
 $ null.value : Named num 0
 $ alternative: chr "two.sided"
 $ method     : chr "Pearson's product-moment correlation"
 $ data.name  : chr "norm_clean$total_protein_mg and norm_clean$activity_mU_per_mg_protein"
 $ conf.int   : num [1:2] -0.98 -0.492

--- norm_clean: clean samples used for the trend fit ---

tibble [8 × 24] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_reps_used               : int [1:8] 3 3 3 3 3 3 3 3
 $ mean_rate_mOD_min         : num [1:8] 7.6 8.5 13.48 10.18 5.62 ...
 $ sd_rate                   : num [1:8] 0.35 0.35 0.293 0.257 0.104 ...
 $ cv_rate                   : num [1:8] 4.61 4.12 2.17 2.52 1.85 ...
 $ family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ individual                : chr [1:8] "01" "02" "03" "04" ...
 $ temperature               : chr [1:8] "ambient" "ambient" "ambient" "ambient" ...
 $ conc_ug_mL                : num [1:8] 876 1039 1900 1129 712 ...
 $ total_protein_mg          : num [1:8] 0.307 0.364 0.665 0.395 0.249 ...
 $ bg_rate_mOD_min           : num [1:8] 0.233 0.233 0.233 0.233 0.233 ...
 $ corrected_mOD_min         : num [1:8] 7.37 8.27 13.25 9.95 5.38 ...
 $ rate_OD_min               : num [1:8] 0.00737 0.00827 0.01325 0.00995 0.00538 ...
 $ nmol_per_min              : num [1:8] 0.228 0.256 0.41 0.308 0.166 ...
 $ activity_mU_per_uL        : num [1:8] 0.1139 0.1278 0.2048 0.1538 0.0832 ...
 $ total_mU_in_homogenate    : num [1:8] 39.9 44.7 71.7 53.8 29.1 ...
 $ activity_mU_per_mg_protein: num [1:8] 130 123 108 136 117 ...
 $ nmol_in_window            : num [1:8] 4.56 5.11 8.19 6.15 3.33 ...
 $ within_std_range          : logi [1:8] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ cv_all                    : num [1:8] 4.61 4.12 2.17 2.52 1.85 ...
 $ n_elevated                : int [1:8] 0 0 0 0 0 0 0 0
 $ n                         : int [1:8] 3 3 3 3 3 3 3 3
 $ cv_flag                   : Factor w/ 3 levels "clean","CV > 15%",..: 1 1 1 1 1 1 1 1
 $ label_n                   : chr [1:8] "3/3 reps" "3/3 reps" "3/3 reps" "3/3 reps" ...

This plate tells a different story from the 20260814 (36 °C) plate: raw activity (mU/µL) rises strongly and significantly with total extracted protein (r = +0.956, p = 0.0002), while mU/mg protein correlates negatively with total extracted protein (r = -0.888, p = 0.0032, n = 8, all 8 samples clean on this plate). Because raw signal genuinely increases with more protein loaded – consistent with more CS enzyme being present when more protein was extracted – the negative mU/mg protein trend here is more consistent with protein normalization doing its intended job (correcting for real inter-sample extraction-efficiency differences) than with a pure division artifact. Protein normalization also compresses the inter-individual spread, from a 2.69-fold raw range down to a 1.64-fold range in mU/mg protein, the opposite direction from the 36 °C plate (where normalizing widened, rather than narrowed, the apparent spread).

10 QC SUMMARY

qc_summary <- build_qc_summary_table(layout_wells, recon$summary, std_curve,
                                      well_diagnostics, well_rates, cv_summary,
                                      baseline_per_sample, pos_control, assay_params)

kable(qc_summary, col.names = c("QC check", "Result"),
      caption = "Quality control summary for this plate")

write.csv(qc_summary, file.path(output_dir, "qc_summary.csv"), row.names = FALSE)

cat("\n--- qc_summary: consolidated QC checks for the run ---\n\n")
str(qc_summary)
QC check Result
Occupied wells in layout 48
Layout wells missing from absorbance CSV 0
Max disagreement between absorbance CSV and full report 0
Standard concentrations with replicate CV > threshold
Standard wells flagged as outliers 0/18 ()
Standard curve R^2 (outlier-excluded, replicate level) 0.9983
Standard curve R^2 (all wells, replicate level) 0.9983
Standard wells that LOST signal over the run 13/18
Standards above photometric linearity ceiling
Background control wells behaving as active reactions 0/3
Sample wells with a decreasing trace 0/24
Sample wells with an elevated starting A412 0/24 ()
Samples with ALL THREE replicates baseline-compromised 0/8 ()
Sample wells with a read glitch 0/24
Sample wells usable for rate extraction 24/24
Samples with technical CV > threshold (all reps) 0/8
Samples with technical CV > threshold (usable reps) 0/8
Positive control replicates rising and linear 3/3

Quality control summary for this plate

--- qc_summary: consolidated QC checks for the run ---

'data.frame':   18 obs. of  2 variables:
 $ check: chr  "Occupied wells in layout" "Layout wells missing from absorbance CSV" "Max disagreement between absorbance CSV and full report" "Standard concentrations with replicate CV > threshold" ...
 $ value: chr  "48" "0" "0" "" ...

10.1 Auto-generated findings

The bullets below are generated programmatically from the QC/CV/anomaly objects above by generate_qc_findings(), so the same logic that flags a problem here is guaranteed to be the logic that flagged it upstream.

qc_findings <- generate_qc_findings(well_diagnostics, well_rates, cv_summary,
                                     baseline_per_sample, std_curve, recon$summary,
                                     assay_params)

cat(paste0("- ", qc_findings, "\n"), sep = "\n")
- **Kinetic sources agree.** absorbance CSV and full report agree exactly on all 528 shared well x timepoint readings; all layout wells are present.

- **Standard signal decay.** 13 of 18 standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.

- **Standard curve fit (outlier-excluded):** slope = 0.03234 A412/nmol, R^2 = 0.9983 (all-wells R^2 = 0.9983).

- **Technical replicate CV.** All 8 samples are within 15% CV across all three replicates.

11 SUMMARY

11.1 Anomalies found

  1. No sample well, positive-control well, or background well is flagged for anything on this plate. No decreasing kinetics, no read glitches, no elevated starting absorbance, and no background well behaving as an active reaction (all three pooled-background replicates are flat).
  2. Widespread but harmless GSH standard signal decay. 13 of 15 (strict drift < 0) to 15 of 18 (including exact-zero ties) non-zero-concentration standard wells lose signal between t = 0 and t = 20 min, consistent with TNB2- instability. This is a larger fraction than on prior plates, but the standard curve is calibrated at t = 0, so it has no effect on the fit (R2 = 0.9983, zero flagged outlier wells).
  3. Elevated positive-control replicate CV. The three positive-control replicates rise linearly (R2 0.997-1.000) but disagree in rate by 22.8% CV (D7 at 2.00 mA412/min vs. D8/D9 at 2.95-3.15 mA412/min), worse precision than any individual sample’s technical triplicate on this plate. This does not affect sample activity values (the positive control is a chemistry validity check only), but is worth watching on future plates.

11.2 Technical replicate precision

All eight samples pass the 15% CV threshold — the highest is F05_01_ambient at 4.6%, and the lowest is F05_08_ambient at 0.8% (F05_02_ambient 4.1%, F05_03_ambient 2.2%, F05_04_ambient 2.5%, F05_05_ambient 1.9%, F05_06_ambient 1.9%, F05_07_ambient 1.3%). No replicate is excluded from any sample’s rate calculation on this plate, so the “all reps” and “usable reps only” CVs are identical throughout.

11.3 Assay validity

The positive control worked in the sense that matters for calibration: all three replicates rise linearly (R2 > 0.997), confirming the core reaction chemistry and reader were functioning. Its elevated replicate-to-replicate CV (22.8%, note above) is a caution rather than a disqualifier, since it does not enter the activity calculation. The GSH standard curve has zero flagged outlier wells and every sample well’s kinetics are clean, so this plate’s activity values carry low calibration uncertainty.

11.4 Recommendation

Treat the activity values here as reliable:

  1. This plate does not need re-assay. No sample is disqualified, no replicate is excluded, and the standard curve has no outlier wells.
  2. Protein normalization looks appropriate for this plate, not artifactual. Raw rate (mU/µL) is strongly and significantly correlated with total extracted protein (r = +0.96, p = 0.0002), the opposite pattern from the 20260814 (36 °C) plate, so the negative mU/mg-protein trend here is better explained by genuine extraction-efficiency differences between samples than by a division artifact. mU/mg protein is therefore a reasonable basis for comparison on this plate.
  3. Watch the positive control’s replicate precision on future plates. A 22.8% CV across a positive-control triplicate is unusually high; if it recurs, it is worth checking pipetting consistency for that specific reagent (e.g., undissolved/settled positive control stock, or a multichannel-pipetting order effect).
  4. The pooled-background design cuts the well count needed for background control roughly 8-fold (3 wells vs. 24 on the paired per-sample design) at the cost of losing any per-sample background estimate; on this plate the background rate is a small fraction of the sample signal (~2.2%) either way, so the simplification appears low-risk here, but a pooled design should be re-validated if background ever becomes a larger fraction of the raw signal.

11.5 What can be used from this plate

All eight samples are usable, and unlike the 36 °C plate, this ambient plate shows real, well-resolved inter-individual variation in CS activity: background-corrected rates range 5.38 to 14.50 mA412/min across the eight F05 individuals at ambient temperature, a 169% spread against 0.8-4.6% technical CV per sample – far larger than the technical noise floor, and in the opposite direction from the essentially uniform activity seen across this same family at 36 °C.

  • Protein-normalized activity (mU/mg protein) is the recommended basis for downstream comparison on this plate, since raw activity correlates strongly with total extracted protein (see note above) – normalizing is removing a real extraction-efficiency covariate here, not introducing an artifact.
  • The elevated positive-control CV (22.8%) is a plate-level caution, not a sample-level one: it does not touch any individual sample’s activity value, but should inform how much weight this plate’s absolute activity values carry relative to future re-assays with tighter positive-control precision.
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase ABCAM https://robertslab.github.io/sams-notebook/posts/2026/2026-08-24-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-5-Ambient-Ctenidia-Re-assay/ Mon, 24 Aug 2026 07:00:00 GMT Citrate Synthase Assay - SORMI June 2026 M.gigas Family 7 Ambient and 36C Ctenidia Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-24-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-7-Ambient-and-36C-Ctenidia/

INTRO

Citrate synthase activity in ctenidia of sixteen Magallana gigas (Pacific oyster) individuals from Family 7 — eight held at ambient temperature and eight at 36°C — assayed together on a single plate using the Citrate Synthase Assay Kit (ABCAM; cat: ab239712). Combining both temperature exposures on one plate removes the between-plate variability that complicates the Family 5 ambient-vs-36°C comparison, where the two exposures were run on separate plates with different read durations. Samples had been previously homogenized on 20260811 (Notebook entry), and protein concentrations for normalization were determined on 20260813 (Notebook entry).

Important

The background control on this plate failed — the reported rates are raw, uncorrected rates.

Reaction Mix (containing CS Substrate) appears to have been added to all three pooled-background wells (F10-F12) instead of Background Control Mix. All three replicates rise substantially over the run (net change 0.067-0.119 A412, vs. <0.05 for a flat background well) rather than staying flat, and this has been confirmed as a pipetting mix-up rather than genuine background signal.

Because 0 of 3 replicates are flat, no background rate can be estimated, and per the pipeline’s existing fallback the background correction applied to every sample on this plate is 0 — i.e. corrected rate = raw rate throughout.

Where a pooled background control has worked correctly, it has been a small term (roughly 2-7% of the sample signal), so the values here are likely only a modest overestimate of true background-corrected activity — but that cannot be confirmed without a working background control. These rates should not be pooled directly with background-corrected values from other assays without accounting for the difference. Re-running just the background triplicate would let this plate’s existing sample, standard-curve, and positive-control data be corrected without re-collecting everything else.

To reduce the risk of this recurring: physically separate the Reaction Mix and Background Control Mix reservoirs/tips, or pipette the background wells first, before switching reagents for the sample wells.

All data/code/analysis is available here:

MATERIALS & METHODS

Citrate Synthase Assay

The Citrate Synthase Assay Kit (ABCAM; cat: ab239712) was used according to the manufacturer’s instructions (ABCAM-Citrate-Synthase-Assay-v4a-ab239712.pdf (GitHub; PDF)). The kit is a coupled kinetic assay: citrate synthase condenses acetyl-CoA and oxaloacetate to release free CoA-SH, the liberated thiol reduces DTNB to TNB2-, and the rate of absorbance increase at 412 nm is proportional to citrate synthase activity. Absolute nmol of thiol are assigned from a reduced glutathione (GSH) standard curve read on the same plate.

Per reaction well, 2 µL of undiluted homogenate was combined with 48 µL of Assay Buffer 7, followed by 50 µL of Reaction Mix. All sixteen samples — eight ambient and eight 36°C — were assayed in technical triplicate on a single clear, flat-bottomed 96-well plate. A six-point GSH standard curve (0, 8, 16, 24, 32, 40 nmol/well) and a CS positive control were included in triplicate on the same plate.

REACTION MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 43 77 3311
DTNB 5 77 385
CS Substrate 2 77 154

BACKGROUND MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 45 4 180
DTNB 5 4 20

Absorbance was read at 412 nm in kinetic mode at 25°C on a Synergy HTX plate reader (Agilent), every 2 minutes for 20 minutes (11 reads).

Normalization

Activity was normalized to total extracted protein rather than tissue weight. Tissue was homogenized in 350 µL of Assay Buffer 7, and protein concentration in each homogenate was quantified on 20260813 (Notebook entry) using the Bradford assay. Total protein mass per sample is that concentration multiplied by the 350 µL homogenization volume. Tissue weights are recorded in the plate layout labels for provenance only and play no role in the activity calculation. Protein values come from two assays run on 20260813: the F07 plate (12 of the 16 samples) and a follow-up re-assay plate (F07_05_36C, F07_01_ambient, F07_04_ambient, F07_06_ambient, re-assayed at dilution after failing QC on the first pass).

Analysis

Data were analyzed in R. The Rmd file is here (GitHub):

Knitted markdown is below.

Background correction was intended to use a single pooled background control rather than a paired per-sample background. A pool was made from 5 µL of each sample homogenate, and 2 µL of that pool (matching the 2 µL sample volume used for every reaction well) was run in triplicate with Background Control Mix (no CS Substrate Mix), producing wells BG-citrate_synthase (F10-F12). As described in the callout above, Reaction Mix was added to these wells instead, so no background correction could be applied.

Data files

Plate reader files

Plate reader files (*.xpt) are available here (require Gen5 software to open):

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/plate_reader_files

Raw absorbance data

Raw absorbance data (*.csv), plate layouts, and Gen5 full reports are available here:

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/raw_absorbance

RESULTS

Apart from the background control, the plate ran well. The positive control was clean (2.6% CV, R2 0.998-1.000) and the GSH standard curve has zero flagged outlier wells (R2 = 0.9992), so calibration uncertainty is low. Fifteen of sixteen samples are fully clean, with every non-failing sample under 3.1% CV on all three raw replicates and most under 2%.

The one failure is A7 (F07_03_ambient), a dead well reading a constant A412 of 0.099 across all 11 timepoints (net change = 0, R2 = 0) rather than a slow reaction. Its two siblings are normal, so F07_03_ambient’s activity is computed from 2 of 3 replicates with a 6.4% usable-replicate CV. Its CV flag column still reads FAIL >15% because that column is deliberately computed on all raw replicates including the dead well — the two figures report different things and both are correct.

Within this plate, raw activity spans 0.0547-0.1356 mU/µL, a real 2.5-fold range against <6.5% technical CV per sample. Protein normalization looks appropriate rather than artifactual: raw rate correlates strongly and significantly with total extracted protein (r = +0.93, p = 0.0000), so mU/mg protein is the recommended basis for downstream comparison.

Sample Reps used CV all reps (%) CV used reps (%) Rate (mA412/min) Activity (mU/uL) Activity (mU/mg protein) CV flag
F07_01_ambient 3/3 1.7 1.7 3.43 0.0547 207.819 pass
F07_02_ambient 3/3 1.9 1.9 7.43 0.1184 109.600 pass
F07_03_ambient 2/3 86.9 6.4 7.18 0.1143 98.879 FAIL >15%
F07_04_ambient 3/3 1.2 1.2 4.97 0.0791 141.243 pass
F07_05_ambient 3/3 1.6 1.6 6.53 0.1040 157.813 pass
F07_06_ambient 3/3 1.8 1.8 7.20 0.1147 102.432 pass
F07_07_ambient 3/3 0.7 0.7 8.22 0.1309 79.566 pass
F07_08_ambient 3/3 0.6 0.6 4.88 0.0778 138.824 pass
F07_01_36C 3/3 1.2 1.2 4.93 0.0786 118.446 pass
F07_02_36C 3/3 1.7 1.7 7.48 0.1192 127.815 pass
F07_03_36C 3/3 3.1 3.1 8.28 0.1319 109.428 pass
F07_04_36C 3/3 1.5 1.5 8.52 0.1356 86.544 pass
F07_05_36C 3/3 0.7 0.7 4.37 0.0695 162.517 pass
F07_06_36C 3/3 2.3 2.3 3.70 0.0589 118.679 pass
F07_07_36C 3/3 0.7 0.7 7.05 0.1123 108.153 pass
F07_08_36C 3/3 0.0 0.0 3.65 0.0581 200.578 pass

All sixteen samples are classified usable. Since both exposures share one plate, one standard curve, and one positive control, the ambient-vs-36°C contrast within Family 7 is not confounded by between-plate differences — though the failed background control means the comparison rests on raw rates.

AI Use Level 4: Substantial AI contribution

1 BACKGROUND

Citrate synthase (CS) activity in ctenidia of sixteen Magallana gigas (Pacific oyster) individuals from family F07 – eight held at ambient temperature and eight at 36 °C – assayed together on a single plate 2026-08-24 with the Abcam Citrate Synthase Assay Kit (ab239712), v4a. Both temperature exposures for F07 are combined on this one plate. This plate also has a confirmed reagent mix-up affecting its background control (note 4) – background-corrected rates on this plate equal the raw rates.

The kit is a coupled kinetic assay. CS condenses acetyl-CoA and oxaloacetate, releasing free CoA-SH; the liberated thiol reduces DTNB to TNB2-, which absorbs at 412 nm. The rate of A412 increase is proportional to CS activity, and absolute nmol of thiol are assigned from a GSH (reduced glutathione) standard curve read on the same plate.

Absorbance was read at 412 nm in kinetic mode, 25 °C, every 2 min for 20 min (11 reads) on a Synergy HTX. See note 1.

Total extracted protein per homogenate (used for normalization, note 2) was not measured on this plate; it comes from two BCA/Bradford-style protein-quantification assays run 2026-08-13: Gen5-20260813-mgig-sormi-BSA-F07-protein.Rmd (12 of the 16 samples) and a follow-up re-assay plate (the remaining 4 – F07_05_36C, F07_01_ambient, F07_04_ambient, F07_06_ambient – which were re-assayed at dilution after failing QC on the first pass). Homogenization used 350 µL of Assay Buffer 7 per sample (reported by the bench operator; see the lab notebook homogenization post for this assay).

This document is fully self-contained: layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are all defined directly below (see ## Pipeline functions under SETUP) rather than sourced from a separate script, so the whole analysis can be reproduced from this one file.

1.1 Sample naming convention

Well labels follow <family>_<individual>_<temperature>-<assay_type>-<weight>-df.<n> for sample wells (e.g. F07_01_ambient-citrate_synthase-7.3-df.0, F07_01_36C-citrate_synthase-9.0-df.0), STD-<assay_type>-<nmol_per_well> for GSH standards, POS-<assay_type> for the positive control, and BG-<assay_type> for the pooled background control (see note 3) – the background label carries no weight or dilution suffix, since it is not tied to any single individual’s tissue. Because both temperature exposures are present on this one plate, <temperature> genuinely varies (ambient or 36C) within this document; nothing in the parsing below hardcodes a temperature. The trailing number in the sample label (e.g. 7.3) is tissue weight in mg, which is not used for normalization here (note 2); it is parsed and validated but plays no role in the activity calculation.

1.2 Notes

  1. Read duration for this plate is 20 minutes (11 reads at 2-min intervals). The assay parameters below set read_duration_min = 20 to match. Rate extraction uses a fixed-width sliding window (rate_window_n = 5 points = 8 min) rather than the full trace; see the KINETIC TRACES section for whether the assay is still in its linear phase over this window.
  2. Activity is normalized to total extracted protein, not tissue weight, because protein concentration was measured directly (BCA-style assay), and controls for extraction efficiency in a way tissue weight cannot. No Sample string in either protein-concentration source needed correction for this plate’s samples.
  3. Background correction uses a single pooled background control, rather than a paired per-sample background for each individual. A pool was made from 5 µL of each sample homogenate, and 2 µL of that pool (matching sample_volume_uL used for every reaction well) was run in triplicate with Background Control Mix (no CS Substrate Mix) to produce wells BG-citrate_synthase (F10-F12). Because there is only one background estimate for the whole plate, the same background-corrected rate would normally be subtracted from every sample – but see note 4.
  4. Reaction Mix (containing CS Substrate) appears to have been added to all three pooled-background wells on this plate instead of Background Control Mix. All three BG-citrate_synthase replicates rise substantially over the run (net change 0.067-0.119 A412, vs. <0.05 for a flat background well) rather than staying flat, and this has been confirmed as a pipetting mix-up rather than genuine background signal. Because 0 of 3 replicates are flat, compute_background_correction() cannot estimate a background rate (bg_rate_flat is NA) and, per its existing fallback, the background correction used for every sample on this plate is 0 – i.e., corrected rate = raw rate throughout. See BACKGROUND CORRECTION and the SUMMARY for what this means for this plate’s results.
library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

1.3 Pipeline functions

All layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are defined directly in this document (rather than sourced from a separate script) so that it is fully self-contained and reproducible from this single file. The functions are grouped below by pipeline stage; each is called in the DATA / KINETIC TRACES / ANOMALY DETECTION / GSH STANDARD CURVE / RATE EXTRACTION / BACKGROUND CORRECTION / TECHNICAL REPLICATE PRECISION / CITRATE SYNTHASE ACTIVITY / QC SUMMARY sections that follow. Nothing plate-specific (typo fixes, disqualified samples, thresholds) is hardcoded in any function – it is always passed in as an argument, from the assay_params list defined further below or from a plate-specific fix list.

1.3.1 Data import and layout parsing

Reads the raw Gen5 exports (kinetic absorbance CSV, full-report text backup, plate-layout CSV), reconciles them against each other, and attaches well-type/sample-id/replicate metadata plus measured protein concentrations to every well.

## -----------------------------------------------------------------------
## elapsed_to_min()
## Purpose : Convert Gen5 "H:MM:SS" elapsed-time strings to decimal minutes.
## Inputs  : x - character vector of "H:MM:SS" strings
## Outputs : numeric vector of elapsed minutes, same length as x
## -----------------------------------------------------------------------
elapsed_to_min <- function(x) {
  vapply(strsplit(x, ":"), function(p) {
    p <- as.numeric(p)
    p[1] * 60 + p[2] + p[3] / 60
  }, numeric(1))
}

## -----------------------------------------------------------------------
## read_absorbance_csv()
## Purpose : Parse a Gen5 "absorbance-*.csv" kinetic export: wavelength
##           header, a "Time," row, a temperature row, then one row per well
##           (well ID in column 1, one OD reading per timepoint thereafter).
## Inputs  : path - path to the absorbance-*.csv file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_absorbance_csv <- function(path) {
  ln <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  time_i   <- grep("^Time,", ln)[1]
  time_str <- strsplit(ln[time_i], ",")[[1]]
  time_str <- time_str[!time_str %in% c("", "Time")]
  tmin     <- elapsed_to_min(time_str)

  well_ln <- ln[grepl("^[A-H][0-9]{1,2},", ln)]
  bind_rows(lapply(strsplit(well_ln, ","), function(p) {
    data.frame(well = p[1], time_min = tmin,
               od = as.numeric(p[seq_along(tmin) + 1]),
               stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## read_full_report()
## Purpose : Parse a Gen5 "full_report-*.txt" export: tab-delimited, wells
##           across columns, timepoints down rows. Used only as an
##           independent cross-check against the absorbance CSV, never as a
##           primary data source.
## Inputs  : path - path to the full_report-*.txt file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_full_report <- function(path) {
  ln  <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  hdr_i <- grep("^Time\t.*\tA1\t", ln)[1]
  hdr   <- trimws(strsplit(ln[hdr_i], "\t")[[1]])
  rows  <- ln[grepl("^[0-9]:[0-9]{2}:[0-9]{2}\t", ln)]
  m     <- do.call(rbind, strsplit(rows, "\t"))
  tmin  <- elapsed_to_min(m[, 1])

  well_cols <- which(grepl("^[A-H][0-9]{1,2}$", hdr))
  bind_rows(lapply(well_cols, function(j) {
    data.frame(well = hdr[j], time_min = tmin,
               od = as.numeric(m[, j]), stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## parse_plate_layout()
## Purpose : Convert a raw Gen5 plate-layout CSV export into one row per
##           occupied well with its descriptive label. Auto-detects two
##           export formats seen in practice:
##             "single"  - one row per plate row (row 1 = column numbers,
##                         column 1 = row letter, last column = "Name" tag);
##                         each cell already holds the full descriptive label.
##             "double"  - each plate row is exported as TWO rows: a
##                         "Well ID" row (SPL codes) immediately followed by
##                         a "Name" row (the descriptive label actually used
##                         downstream).
##           The format is detected from column 1: if a value repeats twice
##           in a row (once for the well-ID row, once for the label row)
##           immediately below a numeric header, it is treated as "double".
## Inputs  : path - path to the layout-*.csv file
## Outputs : data.frame, one row per occupied well:
##           well (chr, e.g. "A1"), plate_row (chr), plate_col (int),
##           label (chr, trimmed descriptive label)
## -----------------------------------------------------------------------
parse_plate_layout <- function(path) {
  plate_layout <- read.csv(path, header = FALSE, stringsAsFactors = FALSE)

  col1 <- trimws(plate_layout$V1)
  # "double" format: row letters appear twice in a row in column 1
  # (once tagging the Well-ID row, once tagging the Name row), e.g.
  # "A", "A", "B", "B", ... Detect by checking whether non-empty,
  # non-header values in col1 are each immediately followed by a repeat.
  data_rows   <- which(col1 != "" & !is.na(col1))[-1]  # drop header row (row 1)
  is_double <- length(data_rows) >= 2 &&
    all(col1[data_rows[c(TRUE, FALSE)]] == col1[data_rows[c(FALSE, TRUE)]][seq_len(floor(length(data_rows) / 2))])

  if (is_double) {
    # Pair each row-letter row with the following label row.
    row_letter_rows <- data_rows[c(TRUE, FALSE)]
    label_rows      <- data_rows[c(FALSE, TRUE)]
    stopifnot(length(row_letter_rows) == length(label_rows))

    out <- expand.grid(pair_idx = seq_along(row_letter_rows),
                        col_idx  = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = col1[row_letter_rows[pair_idx]],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j],
                                   label_rows[pair_idx], col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  } else {
    # "single" format: row 1 = column numbers, column 1 = row letters,
    # last column = "Name" tag; each cell already holds the full label.
    out <- expand.grid(row_idx = 2:nrow(plate_layout),
                        col_idx = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = plate_layout$V1[row_idx],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j], row_idx, col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  }

  attr(out, "layout_format") <- ifelse(is_double, "double", "single")
  out
}

## -----------------------------------------------------------------------
## reconcile_kinetic_sources()
## Purpose : Cross-check the absorbance CSV (primary data source) against
##           the full_report text export (independent instrument export)
##           and the plate layout (expected well set). Fails loudly
##           (stopifnot) if any layout well is missing from the CSV, or if
##           the two kinetic sources disagree on any shared reading.
## Inputs  : absorbance_csv - long-format data.frame from read_absorbance_csv()
##           full_report    - long-format data.frame from read_full_report()
##           layout_wells   - data.frame from parse_plate_layout()
##           tolerance      - max allowed |CSV - report| absolute difference
##                            (default 1e-9, i.e. exact agreement)
## Outputs : list(plate_readings, overlap_check, summary), where
##           plate_readings = absorbance_csv with a `source` column added,
##           summary        = named list of counts for reporting
## -----------------------------------------------------------------------
reconcile_kinetic_sources <- function(absorbance_csv, full_report, layout_wells,
                                       tolerance = 1e-9) {
  wells_csv    <- unique(absorbance_csv$well)
  wells_report <- unique(full_report$well)

  wells_report_not_csv <- setdiff(wells_report, wells_csv)
  wells_layout_not_csv <- setdiff(layout_wells$well, wells_csv)
  wells_csv_not_layout <- setdiff(wells_csv, layout_wells$well)

  overlap_check <- inner_join(absorbance_csv, full_report,
                              by = c("well", "time_min"), suffix = c("_csv", "_report")) %>%
    mutate(abs_diff = abs(od_csv - od_report))

  max_disagreement <- if (nrow(overlap_check) > 0) max(overlap_check$abs_diff) else NA_real_

  stopifnot(
    length(wells_layout_not_csv) == 0,
    is.na(max_disagreement) || max_disagreement < tolerance
  )

  plate_readings <- absorbance_csv %>% mutate(source = "absorbance_csv")

  list(
    plate_readings = plate_readings,
    overlap_check  = overlap_check,
    summary = list(
      n_wells_csv               = length(wells_csv),
      n_wells_report            = length(wells_report),
      n_wells_report_not_csv    = length(wells_report_not_csv),
      n_wells_layout_not_csv    = length(wells_layout_not_csv),
      n_wells_csv_not_layout    = length(wells_csv_not_layout),
      n_shared_readings         = nrow(overlap_check),
      max_disagreement          = max_disagreement
    )
  )
}

## -----------------------------------------------------------------------
## annotate_wells()
## Purpose : Join layout labels onto the kinetic readings and parse each
##           label into well_type, sample_id, family, individual,
##           temperature, tissue weight, and standard concentration.
##           Expected label grammar:
##             sample              : <family>_<individual>_<temperature>-<assay_type>-<weight>-df.<n>
##             standard            : STD-<assay_type>-<nmol_per_well>
##             positive control    : POS-<assay_type>
##             background (pooled) : BG-<assay_type>
##           Unlike the paired per-sample background design used on earlier
##           plates (`_BG` appended to a sample's own label), this plate's
##           background wells are a single pooled control with no sample_id,
##           family, individual, or weight of their own (note 3) -- so
##           sample_id/family/individual/temperature/weight_mg parsing is
##           restricted to well_type == "sample" only.
##           Label typos (e.g. a misspelled assay_type) are corrected before
##           parsing via `label_fixes`, a named character vector of
##           c(pattern = replacement) pairs applied in order with gsub().
## Inputs  : plate_readings - data.frame from reconcile_kinetic_sources()$plate_readings
##           layout_wells   - data.frame from parse_plate_layout()
##           label_fixes    - optional named character vector, e.g.
##                            c("citrate_synthasse" = "citrate_synthase")
## Outputs : plate_long - fully annotated long-format data.frame, one row
##           per well x timepoint, with well_type, sample_id, family,
##           individual, temperature, weight_mg, std_nmol columns added
## -----------------------------------------------------------------------
annotate_wells <- function(plate_readings, layout_wells, label_fixes = NULL) {
  plate_long <- plate_readings %>%
    left_join(layout_wells, by = "well") %>%
    mutate(
      well_type = case_when(
        grepl("^STD-", label) ~ "standard",
        grepl("^POS-", label) ~ "positive_control",
        grepl("^BG-",  label) ~ "background",
        TRUE                  ~ "sample"
      ),
      label_clean = label
    )

  if (!is.null(label_fixes)) {
    for (pat in names(label_fixes)) {
      plate_long$label_clean <- gsub(pat, label_fixes[[pat]], plate_long$label_clean)
    }
  }

  plate_long <- plate_long %>%
    mutate(
      sample_id   = ifelse(well_type == "sample",
                           sub("-[a-zA-Z_]+-.*$", "", label_clean), NA_character_),
      family      = ifelse(!is.na(sample_id), sub("^(F[0-9]+)_.*$", "\\1", sample_id), NA_character_),
      individual  = ifelse(!is.na(sample_id), sub("^F[0-9]+_([0-9]+)_.*$", "\\1", sample_id), NA_character_),
      temperature = ifelse(!is.na(sample_id), sub("^F[0-9]+_[0-9]+_(.*)$", "\\1", sample_id), NA_character_),
      weight_mg   = ifelse(well_type == "sample",
                           as.numeric(sub("^.*-[a-zA-Z_]+-([0-9.]+)-df\\..*$", "\\1", label_clean)),
                           NA_real_),
      std_nmol    = ifelse(well_type == "standard",
                           as.numeric(sub("^STD-[a-zA-Z_]+-", "", label_clean)), NA_real_)
    ) %>%
    arrange(well_type, sample_id, well, time_min)

  # Fail loudly rather than silently dropping malformed labels
  stopifnot(
    !any(is.na(plate_long$well_type)),
    !any(is.na(plate_long$weight_mg[plate_long$well_type == "sample"])),
    !any(is.na(plate_long$std_nmol[plate_long$well_type == "standard"]))
  )

  plate_long
}

## -----------------------------------------------------------------------
## load_protein_concentrations()
## Purpose : Load one or more BCA/Bradford-style protein-concentration CSV
##           exports, concatenate them, and match each plate sample to
##           exactly one protein-concentration record. Fails loudly if any
##           plate sample has zero or more than one match. Converts
##           concentration (ug/mL) to total protein mass (mg) in the full
##           homogenate using the measured homogenization volume.
## Inputs  : protein_files        - character vector of CSV paths, each with
##                                   columns `Sample` and
##                                   `Calculated concentration (ug/mL)`
##           plate_sample_ids     - character vector of this plate's sample IDs
##           homogenate_volume_uL - measured homogenization volume (uL),
##                                   applied identically to every sample
##           sample_id_fixes      - optional named character vector of
##                                   c(pattern = replacement) label-typo
##                                   corrections applied to the protein
##                                   file's Sample column before matching
## Outputs : data.frame, one row per plate sample:
##           sample_id, source_file, conc_ug_mL, total_protein_mg
## -----------------------------------------------------------------------
load_protein_concentrations <- function(protein_files, plate_sample_ids,
                                         homogenate_volume_uL,
                                         sample_id_fixes = NULL) {
  protein_all <- protein_files %>%
    purrr::map_dfr(~ read.csv(.x, check.names = FALSE) %>%
                      mutate(source_file = basename(.x))) %>%
    rename(sample_id_protein = Sample, conc_ug_mL = `Calculated concentration (ug/mL)`) %>%
    select(sample_id_protein, source_file, conc_ug_mL)

  if (!is.null(sample_id_fixes)) {
    for (pat in names(sample_id_fixes)) {
      protein_all$sample_id_protein <- gsub(pat, sample_id_fixes[[pat]],
                                             protein_all$sample_id_protein)
    }
  }

  protein_matches <- protein_all %>% filter(sample_id_protein %in% plate_sample_ids)
  match_counts    <- protein_matches %>% count(sample_id_protein, name = "n_matches")

  stopifnot(
    all(plate_sample_ids %in% protein_matches$sample_id_protein),
    all(match_counts$n_matches == 1)
  )

  protein_matches %>%
    rename(sample_id = sample_id_protein) %>%
    mutate(total_protein_mg = conc_ug_mL * homogenate_volume_uL / 1e6) %>%
    arrange(sample_id)
}

1.3.2 Well-level kinetic QC

Flags read glitches, non-rising (‘decreasing’) traces, elevated t0 baselines, and other well-level anomalies before any rate is trusted.

## -----------------------------------------------------------------------
## local_step_excess()
## Purpose : For a kinetic trace, measure how much each read-to-read step
##           departs from the AVERAGE of its two immediate neighbouring
##           steps. A large local excess flags a read glitch; comparing to
##           local neighbours (rather than the trace's overall median step)
##           tolerates traces that smoothly decelerate throughout.
## Inputs  : od - numeric vector of OD readings in time order
## Outputs : numeric vector of length length(od) - 1 (one per step)
## -----------------------------------------------------------------------
local_step_excess <- function(od) {
  d <- diff(od)
  n <- length(d)
  local_expect <- vapply(seq_len(n), function(i) {
    mean(d[intersect(c(i - 1L, i + 1L), seq_len(n))])
  }, numeric(1))
  d - local_expect
}

## -----------------------------------------------------------------------
## compute_well_diagnostics()
## Purpose : Compute four independent per-well trace-shape diagnostics from
##           the raw kinetic trace, before any rate fitting: direction
##           (net change), monotonicity (fraction of rising steps),
##           over-range (any read above the photometric linearity ceiling),
##           and discontinuity (a read-to-read glitch via local_step_excess).
##           Also flags an elevated starting absorbance for sample wells,
##           and background wells that behave like active reactions.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with od_linear_max, glitch_excess_od,
##                          sample_baseline_od, bg_flat_od_max,
##                          bg_flat_drift_max
## Outputs : well_diagnostics - one row per well, with od_first, od_last,
##           od_max, net_change, frac_rising, max_step, typical_step,
##           step_excess, glitch_at_min, and flag_* / n_flags columns
## -----------------------------------------------------------------------
compute_well_diagnostics <- function(plate_long, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, plate_row, plate_col, well_type, sample_id, label, std_nmol, source) %>%
    summarise(
      od_first        = first(od),
      od_last         = last(od),
      od_max          = max(od),
      net_change      = last(od) - first(od),
      frac_rising     = mean(diff(od) > 0),
      max_step        = diff(od)[which.max(abs(diff(od)))],
      typical_step    = median(abs(diff(od))),
      step_excess     = local_step_excess(od)[which.max(abs(local_step_excess(od)))],
      glitch_at_min   = time_min[which.max(abs(local_step_excess(od))) + 1L],
      .groups = "drop"
    ) %>%
    mutate(
      step_ratio      = abs(max_step) / pmax(typical_step, 0.001),
      flag_decreasing = net_change <= 0,
      flag_over_range = od_max > assay_params$od_linear_max,
      flag_glitch     = abs(step_excess) > assay_params$glitch_excess_od,
      flag_high_baseline = well_type == "sample" & od_first > assay_params$sample_baseline_od,
      flag_bg_active  = well_type == "background" &
                        (od_first > assay_params$bg_flat_od_max |
                         abs(net_change) > assay_params$bg_flat_drift_max),
      n_flags         = flag_decreasing + flag_over_range + flag_glitch +
                        flag_bg_active + flag_high_baseline
    )
}

## -----------------------------------------------------------------------
## compute_baseline_diagnostics()
## Purpose : Classify sample/background wells by starting absorbance
##           (elevated vs. normal) and summarize, per sample, how many of
##           its replicates are baseline-compromised. A sample with all
##           replicates elevated has no clean replicate at all.
## Inputs  : well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with sample_baseline_od
## Outputs : list(baseline_check, baseline_per_sample)
## -----------------------------------------------------------------------
compute_baseline_diagnostics <- function(well_diagnostics, assay_params) {
  baseline_check <- well_diagnostics %>%
    filter(well_type %in% c("sample", "background")) %>%
    mutate(baseline = ifelse(od_first > assay_params$sample_baseline_od,
                             "elevated", "normal")) %>%
    arrange(desc(od_first))

  baseline_per_sample <- baseline_check %>%
    filter(well_type == "sample") %>%
    group_by(sample_id) %>%
    summarise(n_elevated = sum(baseline == "elevated"), n = n(),
              median_baseline = median(od_first), .groups = "drop") %>%
    arrange(desc(n_elevated))

  list(baseline_check = baseline_check, baseline_per_sample = baseline_per_sample)
}

1.3.3 Standard curve and rate extraction

Fits the GSH standard curve, extracts each well’s rate via a sliding-window linear regression gated on R-squared, and summarizes the positive control.

## -----------------------------------------------------------------------
## fit_standard_curve()
## Purpose : Extract t=0 GSH standard readings, quantify signal drift over
##           the run, compute per-concentration replicate statistics, flag
##           outlier wells by deviation from their triplicate median, and
##           fit three candidate calibration lines (all wells, per-
##           concentration means, outlier-excluded). The outlier-excluded
##           fit is the calibration used downstream.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with std_outlier_od
## Outputs : list with standards_t0, standard_drift, standard_summary,
##           standards_flagged, fit_all_wells, fit_means, fit_no_outliers,
##           fit_comparison, std_slope, std_intercept, std_r2, std_nmol_max
## -----------------------------------------------------------------------
fit_standard_curve <- function(plate_long, assay_params) {
  standards_all <- plate_long %>%
    filter(well_type == "standard") %>%
    select(well, std_nmol, time_min, od, source)

  standards_t0 <- standards_all %>%
    filter(time_min == 0) %>%
    select(well, std_nmol, od, source) %>%
    arrange(std_nmol, well)

  standard_drift <- standards_all %>%
    group_by(well, std_nmol) %>%
    summarise(od_t0 = od[time_min == 0], od_t40 = od[time_min == max(time_min)],
              drift = od_t40 - od_t0, .groups = "drop") %>%
    arrange(std_nmol, well)

  standard_summary <- standards_t0 %>%
    group_by(std_nmol) %>%
    summarise(n = n(), mean_od = mean(od), sd_od = sd(od),
              se_od = sd(od) / sqrt(n()), cv_pct = 100 * sd(od) / mean(od),
              median_od = median(od), .groups = "drop") %>%
    arrange(std_nmol) %>%
    mutate(net_od = mean_od - mean_od[std_nmol == 0],
           od_per_nmol = ifelse(std_nmol > 0, net_od / std_nmol, NA_real_))

  standards_flagged <- standards_t0 %>%
    group_by(std_nmol) %>%
    mutate(triplicate_median = median(od),
           deviation = od - triplicate_median,
           is_outlier = abs(deviation) > assay_params$std_outlier_od) %>%
    ungroup() %>%
    arrange(desc(abs(deviation)))

  fit_all_wells   <- lm(od ~ std_nmol, data = standards_flagged)
  fit_means       <- lm(mean_od ~ std_nmol, data = standard_summary)
  fit_no_outliers <- lm(od ~ std_nmol, data = standards_flagged %>% filter(!is_outlier))

  fit_comparison <- bind_rows(
    data.frame(fit = "all wells",          n = nobs(fit_all_wells),
               slope = coef(fit_all_wells)[2],   intercept = coef(fit_all_wells)[1],
               r_squared = summary(fit_all_wells)$r.squared),
    data.frame(fit = "concentration means", n = nobs(fit_means),
               slope = coef(fit_means)[2],       intercept = coef(fit_means)[1],
               r_squared = summary(fit_means)$r.squared),
    data.frame(fit = "outlier-excluded",   n = nobs(fit_no_outliers),
               slope = coef(fit_no_outliers)[2], intercept = coef(fit_no_outliers)[1],
               r_squared = summary(fit_no_outliers)$r.squared)
  )

  list(
    standards_all      = standards_all,
    standards_t0       = standards_t0,
    standard_drift     = standard_drift,
    standard_summary   = standard_summary,
    standards_flagged  = standards_flagged,
    fit_all_wells      = fit_all_wells,
    fit_means          = fit_means,
    fit_no_outliers    = fit_no_outliers,
    fit_comparison     = fit_comparison,
    std_slope          = unname(coef(fit_no_outliers)[2]),
    std_intercept      = unname(coef(fit_no_outliers)[1]),
    std_r2             = summary(fit_no_outliers)$r.squared,
    std_nmol_max       = max(standard_summary$std_nmol)
  )
}

## -----------------------------------------------------------------------
## window_slopes()
## Purpose : Fit every sliding window of `w` consecutive timepoints on one
##           kinetic trace and return the slope (mOD/min) and R^2 of each.
## Inputs  : t - numeric vector of timepoints (min)
##           y - numeric vector of OD readings, same length as t
##           w - window width in number of points
## Outputs : data.frame, one row per window: t_start, t_end,
##           slope_mOD_min, r2
## -----------------------------------------------------------------------
window_slopes <- function(t, y, w) {
  n <- length(y)
  bind_rows(lapply(seq_len(n - w + 1), function(s) {
    i  <- s:(s + w - 1)
    f  <- lm(y[i] ~ t[i])
    r2 <- summary(f)$r.squared
    data.frame(t_start = t[i[1]], t_end = t[i[w]],
               slope_mOD_min = unname(coef(f)[2]) * 1000,
               r2 = ifelse(is.na(r2), 0, r2))
  }))
}

## -----------------------------------------------------------------------
## compute_well_rates()
## Purpose : For every well, scan all sliding windows of assay_params$
##           rate_window_n consecutive reads and record the max-increasing
##           window (used for activity) and the max-absolute window
##           (diagnostic only -- matches Gen5's own `Max V` convention,
##           which can report large negative "rates" for degrading wells).
##           A rate is `rate_usable` only if its R^2 clears the floor, the
##           well's net change is positive, and no read glitch falls inside
##           the fitted window.
## Inputs  : plate_long       - data.frame from annotate_wells()
##           well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with rate_window_n, rate_min_r2
## Outputs : well_rates - one row per well: t_start, t_end, slope_mOD_min,
##           r2, max_abs_slope_mOD_min, abs_window_is_negative,
##           glitch_in_window, rate_usable (plus diagnostic columns joined in)
## -----------------------------------------------------------------------
compute_well_rates <- function(plate_long, well_diagnostics, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, well_type, sample_id, std_nmol) %>%
    group_modify(~ {
      w   <- window_slopes(.x$time_min, .x$od, assay_params$rate_window_n)
      inc <- w[which.max(w$slope_mOD_min), ]
      abs_ <- w[which.max(abs(w$slope_mOD_min)), ]
      data.frame(
        t_start = inc$t_start, t_end = inc$t_end,
        slope_mOD_min = inc$slope_mOD_min, r2 = inc$r2,
        max_abs_slope_mOD_min = abs_$slope_mOD_min,
        abs_window_is_negative = abs_$slope_mOD_min < 0
      )
    }) %>%
    ungroup() %>%
    left_join(well_diagnostics %>% select(well, net_change, frac_rising, od_max,
                                          flag_decreasing, flag_over_range,
                                          flag_glitch, glitch_at_min,
                                          flag_bg_active, n_flags),
              by = "well") %>%
    mutate(
      glitch_in_window = flag_glitch & glitch_at_min > t_start & glitch_at_min <= t_end,
      rate_usable      = r2 >= assay_params$rate_min_r2 & net_change > 0 & !glitch_in_window
    )
}

## -----------------------------------------------------------------------
## compute_positive_control()
## Purpose : Compute the CS Positive Control's rate and activity, and check
##           that its replicates rose linearly -- the plate's proof that the
##           reaction chemistry (DTNB, coupling, reader) worked, independent
##           of any sample-specific problem.
## Inputs  : well_rates   - data.frame from compute_well_rates()
##           std_slope    - calibration slope (A412/nmol), from fit_standard_curve()
##           assay_params - list with sample_volume_uL, dilution_factor
## Outputs : pos_control - one row per positive-control well: t_start, t_end,
##           slope_mOD_min, r2, net_change, activity_mU_uL
## -----------------------------------------------------------------------
compute_positive_control <- function(well_rates, std_slope, assay_params) {
  well_rates %>%
    filter(well_type == "positive_control") %>%
    mutate(activity_mU_uL = (slope_mOD_min / 1000) / std_slope /
                            assay_params$sample_volume_uL * assay_params$dilution_factor)
}

1.3.4 Background correction, replicate CV, and activity calculation

Estimates a single pooled background rate from the flat (well-behaved) background replicates, computes technical-replicate coefficients of variation, and converts corrected rates into protein-normalized CS activity.

## -----------------------------------------------------------------------
## compute_background_correction()
## Purpose : Estimate a single, plate-wide background rate from the flat
##           (well-behaved) replicates of the pooled background control
##           (note 3), since anomalous background wells do not measure
##           background. This plate has one background triplicate for the
##           whole plate, so the estimate is a single row rather than one
##           row per sample_id.
## Inputs  : well_rates - data.frame from compute_well_rates()
## Outputs : background_correction - one-row data.frame: n_bg_total,
##           n_bg_flat, bg_rate_flat (NA if no flat replicate exists),
##           bg_rate_all
## -----------------------------------------------------------------------
compute_background_correction <- function(well_rates) {
  bg <- well_rates %>% filter(well_type == "background")
  data.frame(
    n_bg_total   = nrow(bg),
    n_bg_flat    = sum(!bg$flag_bg_active),
    bg_rate_flat = ifelse(sum(!bg$flag_bg_active) > 0,
                          mean(bg$slope_mOD_min[!bg$flag_bg_active]), NA_real_),
    bg_rate_all  = mean(bg$slope_mOD_min)
  )
}

## -----------------------------------------------------------------------
## compute_replicate_cv()
## Purpose : Compute technical-replicate coefficient of variation (CV%) on
##           the extracted rate, both across ALL three replicates and
##           across USABLE replicates only (rate_usable == TRUE). Flags any
##           sample whose CV exceeds assay_params$cv_threshold_pct.
## Inputs  : well_rates       - data.frame from compute_well_rates()
##           sample_id_order  - character vector giving the row order /
##                              complete sample-ID set to report (typically
##                              the plate's sample IDs, e.g. from
##                              protein_by_sample$sample_id)
##           assay_params     - list with cv_threshold_pct
## Outputs : cv_summary - one row per sample: n_all, mean_all, sd_all,
##           cv_all, n_usable, mean_usable, sd_usable, cv_usable,
##           excluded_wells, fails_cv_all, fails_cv_usable
## -----------------------------------------------------------------------
compute_replicate_cv <- function(well_rates, sample_id_order, assay_params) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  cv_all <- sample_rates %>%
    group_by(sample_id) %>%
    summarise(n_all = n(), mean_all = mean(slope_mOD_min), sd_all = sd(slope_mOD_min),
              cv_all = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  cv_usable <- sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_usable = n(), mean_usable = mean(slope_mOD_min),
              sd_usable = sd(slope_mOD_min),
              cv_usable = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  data.frame(sample_id = sample_id_order, stringsAsFactors = FALSE) %>%
    left_join(cv_all, by = "sample_id") %>%
    left_join(cv_usable, by = "sample_id") %>%
    mutate(
      excluded_wells = vapply(sample_id, function(s) {
        w <- sample_rates$well[sample_rates$sample_id == s & !sample_rates$rate_usable]
        if (length(w) == 0) "-" else paste(w, collapse = ", ")
      }, character(1)),
      fails_cv_all    = cv_all > assay_params$cv_threshold_pct,
      fails_cv_usable = !is.na(cv_usable) & cv_usable > assay_params$cv_threshold_pct
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## calculate_cs_activity()
## Purpose : Compute per-sample CS activity following Abcam sec 10.3, using
##           only usable replicates, corrected for the single pooled
##           background rate (note 3), scaled through the standard curve,
##           and normalized to total extracted protein (rather than tissue
##           weight).
## Inputs  : well_rates            - data.frame from compute_well_rates()
##           protein_by_sample     - data.frame from load_protein_concentrations()
##           background_correction - one-row data.frame from
##                                    compute_background_correction()
##           std_slope             - calibration slope (A412/nmol), from fit_standard_curve()
##           std_nmol_max          - max calibrated GSH concentration, from fit_standard_curve()
##           assay_params          - list with sample_volume_uL, dilution_factor,
##                                    homogenate_volume_uL, read_duration_min
##           plate_long            - data.frame from annotate_wells(), used to
##                                    attach family/individual/temperature
## Outputs : cs_activity - one row per sample: n_reps_used, mean_rate_mOD_min,
##           sd_rate, cv_rate, family, individual, temperature, conc_ug_mL,
##           total_protein_mg, bg_rate_mOD_min, corrected_mOD_min,
##           activity_mU_per_uL, total_mU_in_homogenate,
##           activity_mU_per_mg_protein, within_std_range
## -----------------------------------------------------------------------
calculate_cs_activity <- function(well_rates, protein_by_sample, background_correction,
                                   std_slope, std_nmol_max, assay_params, plate_long) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  bg_rate <- ifelse(is.na(background_correction$bg_rate_flat), 0, background_correction$bg_rate_flat)

  sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_reps_used = n(), mean_rate_mOD_min = mean(slope_mOD_min),
              sd_rate = sd(slope_mOD_min),
              cv_rate = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop") %>%
    left_join(plate_long %>% filter(well_type == "sample") %>%
                distinct(sample_id, family, individual, temperature),
              by = "sample_id") %>%
    left_join(protein_by_sample %>% select(sample_id, conc_ug_mL, total_protein_mg),
              by = "sample_id") %>%
    mutate(
      bg_rate_mOD_min       = bg_rate,
      corrected_mOD_min     = mean_rate_mOD_min - bg_rate_mOD_min,
      rate_OD_min           = corrected_mOD_min / 1000,
      nmol_per_min          = rate_OD_min / std_slope,
      activity_mU_per_uL    = nmol_per_min / assay_params$sample_volume_uL *
                              assay_params$dilution_factor,
      total_mU_in_homogenate = activity_mU_per_uL * assay_params$homogenate_volume_uL,
      activity_mU_per_mg_protein = total_mU_in_homogenate / total_protein_mg,
      nmol_in_window        = nmol_per_min * (assay_params$read_duration_min),
      within_std_range      = nmol_in_window <= std_nmol_max
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## build_results_table()
## Purpose : Assemble the final, presentation-ready per-sample results
##           table from cs_activity, cv_summary and baseline diagnostics,
##           with a plain-language Interpretation column.
## Inputs  : cs_activity          - data.frame from calculate_cs_activity()
##           cv_summary           - data.frame from compute_replicate_cv()
##           baseline_per_sample  - data.frame from compute_baseline_diagnostics()
##           assay_params         - list with cv_threshold_pct, homogenate_volume_uL
## Outputs : results_table - formatted data.frame, ready for kable()/write.csv()
## -----------------------------------------------------------------------
build_results_table <- function(cs_activity, cv_summary, baseline_per_sample, assay_params) {
  cs_activity %>%
    left_join(cv_summary %>% select(sample_id, n_all, cv_all), by = "sample_id") %>%
    left_join(baseline_per_sample %>% select(sample_id, n_elevated), by = "sample_id") %>%
    transmute(
      Sample                       = sample_id,
      Family                       = family,
      Individual                   = individual,
      Temperature                  = temperature,
      `Protein conc (ug/mL)`       = round(conc_ug_mL, 1),
      `Total protein (mg)`         = round(total_protein_mg, 3),
      `Reps used`                  = paste0(n_reps_used, "/", n_all),
      `CV all reps (%)`            = round(cv_all, 1),
      `CV used reps (%)`           = round(cv_rate, 1),
      `Rate (mA412/min)`           = round(mean_rate_mOD_min, 2),
      `BG rate (mA412/min)`        = round(bg_rate_mOD_min, 3),
      `Corrected rate (mA412/min)` = round(corrected_mOD_min, 2),
      `Activity (mU/uL)`           = round(activity_mU_per_uL, 4),
      `Activity (mU/mg protein)`   = round(activity_mU_per_mg_protein, 3),
      `CV flag`                    = ifelse(cv_all > assay_params$cv_threshold_pct,
                                            paste0("FAIL >", assay_params$cv_threshold_pct, "%"), "pass"),
      `Elevated baseline reps`     = paste0(n_elevated, "/3"),
      Interpretation               = case_when(
        n_elevated == 3 ~ "DO NOT USE - no clean replicate",
        n_elevated > 0  ~ "caution - some reps compromised",
        cv_rate > assay_params$cv_threshold_pct | is.na(cv_rate) ~ "caution - imprecise",
        TRUE ~ "usable"
      )
    )
}

1.3.5 QC summary and narrative generation

Assembles the per-well QC summary table and auto-generates the plain-language QC findings bullets used in the QC SUMMARY section below.

## -----------------------------------------------------------------------
## build_qc_summary_table()
## Purpose : Consolidate every QC check performed across the pipeline into a
##           single two-column (check, value) table, suitable for kable() and
##           for writing to qc_summary.csv. Mirrors the individual find
##           functions but as compact, at-a-glance figures rather than prose.
## Inputs  : layout_wells, reconcile_summary, std_curve, well_diagnostics,
##           well_rates, cv_summary, baseline_per_sample, pos_control,
##           assay_params
## Outputs : qc_summary - data.frame with columns check, value (both chr)
## -----------------------------------------------------------------------
build_qc_summary_table <- function(layout_wells, reconcile_summary, std_curve,
                                    well_diagnostics, well_rates, cv_summary,
                                    baseline_per_sample, pos_control, assay_params) {
  standard_summary  <- std_curve$standard_summary
  standards_flagged <- std_curve$standards_flagged
  standard_drift    <- std_curve$standard_drift
  sample_rates      <- well_rates %>% filter(well_type == "sample")

  data.frame(
    check = c(
      "Occupied wells in layout",
      "Layout wells missing from absorbance CSV",
      "Max disagreement between absorbance CSV and full report",
      "Standard concentrations with replicate CV > threshold",
      "Standard wells flagged as outliers",
      "Standard curve R^2 (outlier-excluded, replicate level)",
      "Standard curve R^2 (all wells, replicate level)",
      "Standard wells that LOST signal over the run",
      "Standards above photometric linearity ceiling",
      "Background control wells behaving as active reactions",
      "Sample wells with a decreasing trace",
      "Sample wells with an elevated starting A412",
      "Samples with ALL THREE replicates baseline-compromised",
      "Sample wells with a read glitch",
      "Sample wells usable for rate extraction",
      "Samples with technical CV > threshold (all reps)",
      "Samples with technical CV > threshold (usable reps)",
      "Positive control replicates rising and linear"
    ),
    value = c(
      nrow(layout_wells),
      reconcile_summary$n_wells_layout_not_csv,
      sprintf("%.g", reconcile_summary$max_disagreement),
      paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
            collapse = ", "),
      paste0(sum(standards_flagged$is_outlier), "/", nrow(standards_flagged), " (",
             paste(standards_flagged$well[standards_flagged$is_outlier], collapse = ", "), ")"),
      sprintf("%.4f", std_curve$std_r2),
      sprintf("%.4f", summary(std_curve$fit_all_wells)$r.squared),
      paste0(sum(standard_drift$drift < 0), "/", nrow(standard_drift)),
      paste(standard_summary$std_nmol[standard_summary$mean_od > assay_params$od_linear_max],
            collapse = ", "),
      paste0(sum(well_diagnostics$flag_bg_active), "/",
             sum(well_diagnostics$well_type == "background")),
      paste0(sum(well_diagnostics$flag_decreasing & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(well_diagnostics$flag_high_baseline), "/",
             sum(well_diagnostics$well_type == "sample"), " (",
             paste(well_diagnostics$well[well_diagnostics$flag_high_baseline], collapse = ", "), ")"),
      paste0(sum(baseline_per_sample$n_elevated == baseline_per_sample$n), "/",
             nrow(baseline_per_sample), " (",
             paste(baseline_per_sample$sample_id[baseline_per_sample$n_elevated == baseline_per_sample$n],
                   collapse = ", "), ")"),
      paste0(sum(well_diagnostics$flag_glitch & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(sample_rates$rate_usable), "/", nrow(sample_rates)),
      paste0(sum(cv_summary$fails_cv_all), "/", nrow(cv_summary)),
      paste0(sum(cv_summary$fails_cv_usable), "/", nrow(cv_summary)),
      paste0(sum(pos_control$net_change > 0 & pos_control$r2 > 0.99), "/", nrow(pos_control))
    ),
    stringsAsFactors = FALSE
  )
}

## -----------------------------------------------------------------------
## generate_qc_findings()
## Purpose : Walk the QC/CV/anomaly objects for a plate and emit a
##           consistent, generic-language markdown bullet list of findings
##           (decreasing/glitched/high-baseline wells, background-control
##           mix-ups, standard-curve QC failures, CV failures, disqualified
##           samples). Intended to seed the SUMMARY section of a plate's
##           .Rmd; a short hand-written subsection for plate-specific
##           interpretation should follow it.
## Inputs  : well_diagnostics    - data.frame from compute_well_diagnostics()
##           well_rates          - data.frame from compute_well_rates()
##           cv_summary          - data.frame from compute_replicate_cv()
##           baseline_per_sample - data.frame from compute_baseline_diagnostics()
##           std_curve           - list from fit_standard_curve()
##           reconcile_summary   - list from reconcile_kinetic_sources()$summary
##           assay_params        - list with cv_threshold_pct, od_linear_max
## Outputs : character vector, one markdown bullet per finding (empty
##           findings are omitted, so a clean plate returns a short list)
## -----------------------------------------------------------------------
generate_qc_findings <- function(well_diagnostics, well_rates, cv_summary,
                                  baseline_per_sample, std_curve,
                                  reconcile_summary, assay_params) {
  bullets <- character(0)

  # Source reconciliation
  if (reconcile_summary$n_wells_layout_not_csv > 0 ||
      (!is.na(reconcile_summary$max_disagreement) && reconcile_summary$max_disagreement >= 1e-9)) {
    bullets <- c(bullets, sprintf(
      "**Kinetic source disagreement.** %d layout well(s) missing from the absorbance CSV; max disagreement between absorbance CSV and full report was %.4g A412.",
      reconcile_summary$n_wells_layout_not_csv, reconcile_summary$max_disagreement))
  } else {
    bullets <- c(bullets, sprintf(
      "**Kinetic sources agree.** absorbance CSV and full report agree exactly on all %d shared well x timepoint readings; all layout wells are present.",
      reconcile_summary$n_shared_readings))
  }

  # Sample wells with decreasing kinetics
  dec_samples <- well_diagnostics %>% filter(well_type == "sample", flag_decreasing)
  if (nrow(dec_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Decreasing kinetics.** %d sample well(s) end lower than (or equal to) their starting A412 and are excluded from rate extraction: %s.",
      nrow(dec_samples), paste(dec_samples$well, collapse = ", ")))
  }

  # Read glitches
  glitch_samples <- well_diagnostics %>% filter(well_type == "sample", flag_glitch)
  if (nrow(glitch_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Read glitches.** %d sample well(s) contain a single read-to-read step inconsistent with its neighbours: %s (only disqualifying if it falls inside the fitted rate window).",
      nrow(glitch_samples), paste(glitch_samples$well, collapse = ", ")))
  }

  # Elevated baseline
  high_baseline <- well_diagnostics %>% filter(well_type == "sample", flag_high_baseline)
  if (nrow(high_baseline) > 0) {
    bullets <- c(bullets, sprintf(
      "**Elevated starting absorbance.** %d sample well(s) start above the baseline threshold (A412 > %s), indicating pre-existing thiol/contamination before the reaction began: %s.",
      nrow(high_baseline), assay_params$sample_baseline_od, paste(high_baseline$well, collapse = ", ")))
  }

  # Samples fully disqualified by baseline
  disqualified <- baseline_per_sample %>% filter(n_elevated == n)
  if (nrow(disqualified) > 0) {
    bullets <- c(bullets, sprintf(
      "**Sample(s) with no clean replicate.** Every replicate of %s has an elevated starting A412; these samples have no usable measurement on this plate regardless of CV.",
      paste(disqualified$sample_id, collapse = ", ")))
  }

  # Over-range wells
  over_range <- well_diagnostics %>% filter(flag_over_range)
  if (nrow(over_range) > 0) {
    bullets <- c(bullets, sprintf(
      "**Over-range wells.** %d well(s) exceed the photometric linearity ceiling (A412 > %s): %s.",
      nrow(over_range), assay_params$od_linear_max, paste(over_range$well, collapse = ", ")))
  }

  # Background wells behaving as active reactions
  bg_active <- well_diagnostics %>% filter(well_type == "background", flag_bg_active)
  if (nrow(bg_active) > 0) {
    bullets <- c(bullets, sprintf(
      "**Background control wells behaving as active reactions.** %d of %d background wells show a starting A412 or drift inconsistent with a flat, inactive well: %s.",
      nrow(bg_active), sum(well_diagnostics$well_type == "background"),
      paste(bg_active$well, collapse = ", ")))
  }

  # Standard curve QC
  std_summary <- std_curve$standard_summary
  cv_fail_std <- std_summary %>% filter(cv_pct > assay_params$cv_threshold_pct)
  if (nrow(cv_fail_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**GSH standard curve replicate imprecision.** %d of %d concentrations exceed %s%% CV: %s nmol/well.",
      nrow(cv_fail_std), nrow(std_summary), assay_params$cv_threshold_pct,
      paste(cv_fail_std$std_nmol, collapse = ", ")))
  }
  n_std_outliers <- sum(std_curve$standards_flagged$is_outlier)
  if (n_std_outliers > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard outlier wells.** %d of %d standard wells deviate from their triplicate median by more than %s A412: %s.",
      n_std_outliers, nrow(std_curve$standards_flagged), assay_params$std_outlier_od,
      paste(std_curve$standards_flagged$well[std_curve$standards_flagged$is_outlier], collapse = ", ")))
  }
  over_range_std <- std_summary %>% filter(mean_od > assay_params$od_linear_max)
  if (nrow(over_range_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**Standards above the photometric linear range.** %s nmol/well read above A412 %s; the calibration is trustworthy only below these concentrations.",
      paste(over_range_std$std_nmol, collapse = ", "), assay_params$od_linear_max))
  }
  n_drift_neg <- sum(std_curve$standard_drift$drift < 0)
  if (n_drift_neg > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard signal decay.** %d of %d standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.",
      n_drift_neg, nrow(std_curve$standard_drift)))
  }
  bullets <- c(bullets, sprintf(
    "**Standard curve fit (outlier-excluded):** slope = %.5f A412/nmol, R^2 = %.4f (all-wells R^2 = %.4f).",
    std_curve$std_slope, std_curve$std_r2, summary(std_curve$fit_all_wells)$r.squared))

  # Technical replicate CV failures
  fail_all <- cv_summary %>% filter(fails_cv_all)
  if (nrow(fail_all) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% (all replicates).** %d of %d samples: %s.",
      assay_params$cv_threshold_pct, nrow(fail_all), nrow(cv_summary),
      paste(sprintf("%s (%.0f%%)", fail_all$sample_id, fail_all$cv_all), collapse = ", ")))
  } else {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV.** All %d samples are within %s%% CV across all three replicates.",
      nrow(cv_summary), assay_params$cv_threshold_pct))
  }
  fail_usable <- cv_summary %>% filter(fails_cv_usable)
  if (nrow(fail_usable) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% persists after excluding flagged replicates.** %s.",
      assay_params$cv_threshold_pct,
      paste(sprintf("%s (%.0f%%, n=%d)", fail_usable$sample_id, fail_usable$cv_usable,
                    fail_usable$n_usable), collapse = ", ")))
  }

  bullets
}

1.4 Assay parameters

All assay constants and QC thresholds are collected here so nothing downstream hard-codes a number.

assay_params <- list(
  sample_volume_uL      = 2,     # homogenate volume per reaction well (V in Abcam formula); also the pooled-background volume (note 3)
  buffer_volume_uL      = 48,    # Assay Buffer 7 added to reach 50 uL pre-Reaction Mix
  dilution_factor       = 1,     # D; layout `df.0` = undiluted homogenate
  homogenate_volume_uL  = 350,   # MEASURED buffer volume tissue was homogenized in (see note 2)
  read_wavelength_nm    = 412,
  read_interval_min     = 2,
  read_duration_min     = 20,    # see note 1
  rate_window_n         = 5,     # points per sliding regression window (5 pts = 8 min)
  rate_min_r2           = 0.80,  # min R^2 for a rate window to be trusted
  glitch_excess_od      = 0.02,  # |step - mean of neighbouring steps| flagging a read glitch
  cv_threshold_pct      = 15,    # technical-replicate CV QC threshold
  od_linear_max         = 1.5,   # upper A412 bound for reliable photometry
  std_outlier_od        = 0.15,  # |deviation from triplicate median| flagging a standard well
  sample_baseline_od    = 0.35,  # max acceptable t0 A412 for a sample well
  bg_flat_od_max        = 0.15,  # a well-behaved background well starts below this A412
  bg_flat_drift_max     = 0.05   # ...and drifts less than this over the whole run
)

cat("--- assay_params: assay constants and QC thresholds ---\n\n")
str(assay_params)
--- assay_params: assay constants and QC thresholds ---

List of 16
 $ sample_volume_uL    : num 2
 $ buffer_volume_uL    : num 48
 $ dilution_factor     : num 1
 $ homogenate_volume_uL: num 350
 $ read_wavelength_nm  : num 412
 $ read_interval_min   : num 2
 $ read_duration_min   : num 20
 $ rate_window_n       : num 5
 $ rate_min_r2         : num 0.8
 $ glitch_excess_od    : num 0.02
 $ cv_threshold_pct    : num 15
 $ od_linear_max       : num 1.5
 $ std_outlier_od      : num 0.15
 $ sample_baseline_od  : num 0.35
 $ bg_flat_od_max      : num 0.15
 $ bg_flat_drift_max   : num 0.05

1.5 Output directory

# Output directory for this analysis (matches this file's name, per ../code/README.md)
output_dir <- "../outputs/Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- output_dir: destination for all figures and tables ---\n")
str(output_dir)
--- output_dir: destination for all figures and tables ---
 chr "../outputs/Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C"

2 DATA

Data are read from the local repo (../data/raw_absorbance/) so this document renders before/after the files are pushed to GitHub.

2.1 Plate layout

data_dir <- "../data/raw_absorbance"
run_stem <- "Gen5-20260824-mgig-sormi-citrate_synthase-F07-ambient_and_36C"

layout_wells <- parse_plate_layout(file.path(data_dir, paste0("layout-", run_stem, ".csv")))

cat("Layout format detected:", attr(layout_wells, "layout_format"), "\n")
cat("Occupied wells in layout:", nrow(layout_wells), "\n")

cat("\n--- layout_wells: one row per occupied well, with its descriptive label ---\n\n")
str(layout_wells)
Layout format detected: single 
Occupied wells in layout: 72 

--- layout_wells: one row per occupied well, with its descriptive label ---

'data.frame':   72 obs. of  4 variables:
 $ well     : chr  "A1" "A2" "A3" "A4" ...
 $ plate_row: chr  "A" "A" "A" "A" ...
 $ plate_col: int  1 2 3 4 5 6 7 8 9 10 ...
 $ label    : chr  "F07_01_ambient-citrate_synthase-7.3-df.0" "F07_01_ambient-citrate_synthase-7.3-df.0" "F07_01_ambient-citrate_synthase-7.3-df.0" "F07_02_ambient-citrate_synthase-19.4-df.0" ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "row_idx" "col_idx"
  ..$ dimnames:List of 2
  .. ..$ row_idx: chr [1:8] "row_idx=2" "row_idx=3" "row_idx=4" "row_idx=5" ...
  .. ..$ col_idx: chr [1:12] "col_idx= 2" "col_idx= 3" "col_idx= 4" "col_idx= 5" ...
 - attr(*, "layout_format")= chr "single"

2.2 Kinetic readers

absorbance_csv <- read_absorbance_csv(file.path(data_dir, paste0("absorbance-", run_stem, ".csv")))
full_report    <- read_full_report(file.path(data_dir, paste0("full_report-", run_stem, ".txt")))

cat("absorbance CSV : ", dplyr::n_distinct(absorbance_csv$well), " wells x ",
    dplyr::n_distinct(absorbance_csv$time_min), " timepoints\n", sep = "")
cat("full report    : ", dplyr::n_distinct(full_report$well), " wells x ",
    dplyr::n_distinct(full_report$time_min), " timepoints\n", sep = "")

cat("\n--- absorbance_csv: long-format readings from absorbance-*.csv ---\n\n")
str(absorbance_csv)
cat("\n--- full_report: long-format readings from full_report-*.txt ---\n\n")
str(full_report)
absorbance CSV : 72 wells x 11 timepoints
full report    : 72 wells x 11 timepoints

--- absorbance_csv: long-format readings from absorbance-*.csv ---

'data.frame':   792 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.114 0.12 0.127 0.134 0.141 0.147 0.154 0.161 0.167 0.173 ...

--- full_report: long-format readings from full_report-*.txt ---

'data.frame':   792 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.114 0.12 0.127 0.134 0.141 0.147 0.154 0.161 0.167 0.173 ...

2.3 Cross-check against the full report

absorbance-*.csv is used for every well; full_report-*.txt is used only to confirm the CSV is trustworthy: every well it reports should also be in the CSV, every well the layout expects should be in the CSV, and every reading the two files share should agree exactly.

recon <- reconcile_kinetic_sources(absorbance_csv, full_report, layout_wells)
plate_readings <- recon$plate_readings

cat("Wells in absorbance CSV          :", recon$summary$n_wells_csv, "\n")
cat("Wells in full report             :", recon$summary$n_wells_report, "\n")
cat("Wells in full report not in CSV  :", recon$summary$n_wells_report_not_csv, "\n")
cat("Layout wells missing from CSV    :", recon$summary$n_wells_layout_not_csv, "\n")
cat("CSV wells not in layout          :", recon$summary$n_wells_csv_not_layout, "\n")
cat("Shared well x timepoints         :", recon$summary$n_shared_readings, "\n")
cat("Max |CSV - report| disagreement  :", recon$summary$max_disagreement, "\n")

cat("\n--- recon$overlap_check: per-reading comparison of the two raw files ---\n\n")
str(recon$overlap_check)
cat("\n--- plate_readings: readings for all occupied wells, from absorbance_csv ---\n\n")
str(plate_readings)
Wells in absorbance CSV          : 72 
Wells in full report             : 72 
Wells in full report not in CSV  : 0 
Layout wells missing from CSV    : 0 
CSV wells not in layout          : 0 
Shared well x timepoints         : 792 
Max |CSV - report| disagreement  : 0 

--- recon$overlap_check: per-reading comparison of the two raw files ---

'data.frame':   792 obs. of  5 variables:
 $ well     : chr  "A1" "A1" "A1" "A1" ...
 $ time_min : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od_csv   : num  0.114 0.12 0.127 0.134 0.141 0.147 0.154 0.161 0.167 0.173 ...
 $ od_report: num  0.114 0.12 0.127 0.134 0.141 0.147 0.154 0.161 0.167 0.173 ...
 $ abs_diff : num  0 0 0 0 0 0 0 0 0 0 ...

--- plate_readings: readings for all occupied wells, from absorbance_csv ---

'data.frame':   792 obs. of  4 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.114 0.12 0.127 0.134 0.141 0.147 0.154 0.161 0.167 0.173 ...
 $ source  : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...

This plate’s absorbance-*.csv covers the full 72 occupied wells (48 sample + 18 standard + 3 positive-control + 3 pooled-background, note 3) and the layout carries no spelling typo, so no label fix is needed here. An earlier export of this file was missing 6 wells (the BG triplicate and the STD-0 triplicate) relative to the layout and full_report; it was re-exported from Gen5 before this analysis was run, and the check above now passes cleanly.

2.4 Annotate wells and parse metadata

plate_long <- suppressWarnings(annotate_wells(plate_readings, layout_wells))

cat("Wells per type:\n")
print(plate_long %>% distinct(well, well_type) %>% count(well_type, name = "n_wells") %>%
        as.data.frame(), row.names = FALSE)
cat("\nSamples:", paste(sort(unique(na.omit(plate_long$sample_id))), collapse = ", "), "\n")
cat("Standards (nmol/well):", paste(sort(unique(na.omit(plate_long$std_nmol))), collapse = ", "), "\n")

cat("\n--- plate_long: fully annotated long-format plate, one row per well x timepoint ---\n\n")
str(plate_long)
Wells per type:
        well_type n_wells
       background       3
 positive_control       3
           sample      48
         standard      18

Samples: F07_01_36C, F07_01_ambient, F07_02_36C, F07_02_ambient, F07_03_36C, F07_03_ambient, F07_04_36C, F07_04_ambient, F07_05_36C, F07_05_ambient, F07_06_36C, F07_06_ambient, F07_07_36C, F07_07_ambient, F07_08_36C, F07_08_ambient 
Standards (nmol/well): 0, 8, 16, 24, 32, 40 

--- plate_long: fully annotated long-format plate, one row per well x timepoint ---

'data.frame':   792 obs. of  15 variables:
 $ well       : chr  "F10" "F10" "F10" "F10" ...
 $ time_min   : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od         : num  0.156 0.176 0.196 0.213 0.227 0.239 0.249 0.257 0.263 0.269 ...
 $ source     : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ plate_row  : chr  "F" "F" "F" "F" ...
 $ plate_col  : int  10 10 10 10 10 10 10 10 10 10 ...
 $ label      : chr  "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" ...
 $ well_type  : chr  "background" "background" "background" "background" ...
 $ label_clean: chr  "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" "BG-citrate_synthase" ...
 $ sample_id  : chr  NA NA NA NA ...
 $ family     : chr  NA NA NA NA ...
 $ individual : chr  NA NA NA NA ...
 $ temperature: chr  NA NA NA NA ...
 $ weight_mg  : num  NA NA NA NA NA NA NA NA NA NA ...
 $ std_nmol   : num  NA NA NA NA NA NA NA NA NA NA ...

annotate_wells() emits a harmless NAs introduced by coercion warning (suppressed above), because its weight_mg parser only applies to sample wells and returns NA for standards/positive control/background by construction (background is a pooled control with no individual weight of its own, note 3). No weight_mg value is actually missing for a sample well; annotate_wells()’s internal stopifnot() would halt otherwise.

2.5 Protein concentration (normalization factor)

Activity is normalized to total extracted protein rather than tissue weight (note 2). Protein concentration for each sample was measured on a separate BCA/Bradford-style plate (595 nm): 12 of this plate’s 16 samples come from the primary F07 protein-quantification run, and the remaining 4 (flagged for re-assay in that run) come from a follow-up re-assay plate. No Sample string typo needed correction for this plate’s samples.

protein_files <- c(
  "../data/BSA/raw_absorbance/sample_protein_concentrations_2.csv",
  "../outputs/Gen5-20260813-mgig-BSA-F07/sample_protein_concentrations.csv"
)

plate_sample_ids <- sort(unique(na.omit(plate_long$sample_id)))

protein_by_sample <- load_protein_concentrations(protein_files, plate_sample_ids,
                                                  assay_params$homogenate_volume_uL) %>%
  left_join(plate_long %>% filter(well_type == "sample") %>%
              distinct(sample_id, family, individual, temperature),
            by = "sample_id")

cat("Samples matched to a protein concentration record:", nrow(protein_by_sample),
    "/", length(plate_sample_ids), "\n")
cat("Protein concentration range (ug/mL):",
    paste(range(protein_by_sample$conc_ug_mL), collapse = " - "), "\n")
cat("Total protein per homogenate range (mg):",
    paste(round(range(protein_by_sample$total_protein_mg), 3), collapse = " - "), "\n\n")
print(protein_by_sample %>% select(sample_id, source_file, conc_ug_mL, total_protein_mg) %>%
        as.data.frame(), row.names = FALSE)

cat("\n--- protein_by_sample: matched protein concentration and total protein per homogenate ---\n\n")
str(protein_by_sample)
Samples matched to a protein concentration record: 16 / 16 
Protein concentration range (ug/mL): 263.1 - 1644.6 
Total protein per homogenate range (mg): 0.092 - 0.576 

      sample_id                         source_file conc_ug_mL total_protein_mg
     F07_01_36C   sample_protein_concentrations.csv      663.3         0.232155
 F07_01_ambient sample_protein_concentrations_2.csv      263.1         0.092085
     F07_02_36C   sample_protein_concentrations.csv      932.4         0.326340
 F07_02_ambient   sample_protein_concentrations.csv     1080.1         0.378035
     F07_03_36C   sample_protein_concentrations.csv     1205.5         0.421925
 F07_03_ambient   sample_protein_concentrations.csv     1155.6         0.404460
     F07_04_36C   sample_protein_concentrations.csv     1567.2         0.548520
 F07_04_ambient sample_protein_concentrations_2.csv      560.0         0.196000
     F07_05_36C sample_protein_concentrations_2.csv      427.9         0.149765
 F07_05_ambient   sample_protein_concentrations.csv      659.3         0.230755
     F07_06_36C   sample_protein_concentrations.csv      496.5         0.173775
 F07_06_ambient sample_protein_concentrations_2.csv     1119.4         0.391790
     F07_07_36C   sample_protein_concentrations.csv     1038.1         0.363335
 F07_07_ambient   sample_protein_concentrations.csv     1644.6         0.575610
     F07_08_36C   sample_protein_concentrations.csv      289.8         0.101430
 F07_08_ambient   sample_protein_concentrations.csv      560.2         0.196070

--- protein_by_sample: matched protein concentration and total protein per homogenate ---

'data.frame':   16 obs. of  7 variables:
 $ sample_id       : chr  "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ source_file     : chr  "sample_protein_concentrations.csv" "sample_protein_concentrations_2.csv" "sample_protein_concentrations.csv" "sample_protein_concentrations.csv" ...
 $ conc_ug_mL      : num  663 263 932 1080 1206 ...
 $ total_protein_mg: num  0.2322 0.0921 0.3263 0.378 0.4219 ...
 $ family          : chr  "F07" "F07" "F07" "F07" ...
 $ individual      : chr  "01" "01" "02" "02" ...
 $ temperature     : chr  "36C" "ambient" "36C" "ambient" ...

3 KINETIC TRACES

Every trace is plotted before any rate is extracted. In this assay the shape of the trace is the primary diagnostic: CS activity must produce a monotonically rising A412, and only a genuinely linear stretch may be used for a rate.

trace_plot <- ggplot(plate_long, aes(x = time_min, y = od, group = well, colour = well_type)) +
  geom_line(linewidth = 0.5) +
  geom_hline(yintercept = assay_params$od_linear_max,
             linetype = "dotted", colour = "grey40") +
  facet_wrap(~ plate_row + plate_col, ncol = 12,
             labeller = function(d) list(paste0(d$plate_row, d$plate_col))) +
  scale_colour_manual(values = c(sample = "#08519c", background = "#a63603",
                                 standard = "#238b45", positive_control = "#6a51a3")) +
  labs(title = "A412 kinetic traces, all occupied wells",
       subtitle = paste0("Dotted line = assumed photometric linearity ceiling (A412 = ",
                         assay_params$od_linear_max,
                         "). Rising traces only are usable for rate extraction."),
       x = "Time (min)", y = "A412", colour = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        strip.text = element_text(size = 6),
        axis.text  = element_text(size = 5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "kinetic_traces_all_wells.png"), trace_plot,
       width = 11, height = 8, dpi = 300)

cat("--- trace_plot: ggplot object structure ---\n\n")
summary(trace_plot)

trace_plot

--- trace_plot: ggplot object structure ---

data: well, time_min, od, source, plate_row, plate_col, label,
  well_type, label_clean, sample_id, family, individual, temperature,
  weight_mg, std_nmol [792x15]
mapping:  x = ~time_min, y = ~od, group = ~well, colour = ~well_type
scales:   colour 
faceting:  ~plate_row, ~plate_col 
-----------------------------------
geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

Nearly every sample and positive-control trace on this plate rises across the full 20-minute window (replicate-level R2 > 0.99 for the fitted rate window on 47 of the 48 sample wells, computed below) – the sole exception, A7 (F07_03_ambient), is completely flat at A412 0.099 for all 11 reads, a dead/failed well rather than a slow reaction (see ANOMALY DETECTION). The three pooled-background traces (F10-F12) are not flat here – they rise at a rate closer to a sample well than a background well, the signature of the confirmed Reaction Mix mix-up (note 4).

4 ANOMALY DETECTION

4.1 Per-well trace diagnostics

Four independent checks, computed from the raw trace before any rate fitting:

  1. Direction — net A412 change from first to last read. A CS-containing well must be positive.
  2. Monotonicity — fraction of the 10 read-to-read intervals that rise.
  3. Over-range — any read above the photometric linearity ceiling.
  4. Discontinuity — a single read-to-read step that departs from its two neighbouring steps, catching read glitches rather than biology.

Two further checks specific to well role: an elevated starting A412 (pre-existing thiol/contamination) for sample wells, and a background well whose trace is not flat (indicating it received Reaction Mix rather than Background Control Mix).

well_diagnostics <- compute_well_diagnostics(plate_long, assay_params)

cat("Flag counts across all", nrow(well_diagnostics), "wells:\n")
print(well_diagnostics %>% summarise(across(starts_with("flag_"), sum)) %>%
        as.data.frame(), row.names = FALSE)

cat("\nWells with at least one flag:", sum(well_diagnostics$n_flags > 0), "/", nrow(well_diagnostics), "\n")

cat("\n--- well_diagnostics: per-well trace diagnostics and anomaly flags ---\n\n")
str(well_diagnostics)
Flag counts across all 72 wells:
 flag_decreasing flag_over_range flag_glitch flag_high_baseline flag_bg_active
              16               0           0                  0              3

Wells with at least one flag: 19 / 72 

--- well_diagnostics: per-well trace diagnostics and anomaly flags ---

tibble [72 × 24] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:72] "A1" "A10" "A11" "A12" ...
 $ plate_row         : chr [1:72] "A" "A" "A" "A" ...
 $ plate_col         : int [1:72] 1 10 11 12 2 3 4 5 6 7 ...
 $ well_type         : chr [1:72] "sample" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:72] "F07_01_ambient" "F07_04_ambient" "F07_04_ambient" "F07_04_ambient" ...
 $ label             : chr [1:72] "F07_01_ambient-citrate_synthase-7.3-df.0" "F07_04_ambient-citrate_synthase-10.7-df.0" "F07_04_ambient-citrate_synthase-10.7-df.0" "F07_04_ambient-citrate_synthase-10.7-df.0" ...
 $ std_nmol          : num [1:72] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:72] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:72] 0.114 0.126 0.126 0.126 0.115 0.115 0.147 0.148 0.146 0.099 ...
 $ od_last           : num [1:72] 0.18 0.22 0.22 0.221 0.182 0.181 0.263 0.268 0.263 0.099 ...
 $ od_max            : num [1:72] 0.18 0.22 0.22 0.221 0.182 0.181 0.263 0.268 0.263 0.099 ...
 $ net_change        : num [1:72] 0.066 0.094 0.094 0.095 0.067 0.066 0.116 0.12 0.117 0 ...
 $ frac_rising       : num [1:72] 1 1 1 1 1 1 1 1 1 0 ...
 $ max_step          : num [1:72] 0.007 0.01 0.01 0.01 0.007 0.007 0.016 0.016 0.016 0 ...
 $ typical_step      : num [1:72] 0.007 0.0095 0.0095 0.0095 0.007 ...
 $ step_excess       : num [1:72] 0.001 0.0015 -0.001 -0.001 -0.001 ...
 $ glitch_at_min     : num [1:72] 20 14 18 10 18 2 2 6 2 2 ...
 $ step_ratio        : num [1:72] 1 1.05 1.05 1.05 1 ...
 $ flag_decreasing   : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range   : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:72] 0 0 0 0 0 0 0 0 0 1 ...

4.2 Anomaly table

anomaly_table <- well_diagnostics %>%
  filter(n_flags > 0) %>%
  select(well, sample_id, well_type, od_first, net_change, n_flags,
         flag_decreasing, flag_over_range, flag_glitch,
         flag_high_baseline, flag_bg_active) %>%
  arrange(desc(n_flags), well)

kable(anomaly_table, digits = 3,
      caption = "Wells with at least one anomaly flag")

cat("\nTotal flagged wells:", nrow(anomaly_table), "/", nrow(well_diagnostics), "\n")

cat("\n--- anomaly_table: all flagged wells with their specific flags ---\n\n")
str(anomaly_table)
well sample_id well_type od_first net_change n_flags flag_decreasing flag_over_range flag_glitch flag_high_baseline flag_bg_active
A7 F07_03_ambient sample 0.099 0.000 1 TRUE FALSE FALSE FALSE FALSE
E10 NA standard 0.849 -0.013 1 TRUE FALSE FALSE FALSE FALSE
E11 NA standard 0.854 -0.011 1 TRUE FALSE FALSE FALSE FALSE
E12 NA standard 0.856 -0.013 1 TRUE FALSE FALSE FALSE FALSE
E4 NA standard 0.364 -0.004 1 TRUE FALSE FALSE FALSE FALSE
E5 NA standard 0.373 -0.004 1 TRUE FALSE FALSE FALSE FALSE
E6 NA standard 0.373 -0.003 1 TRUE FALSE FALSE FALSE FALSE
E7 NA standard 0.636 -0.007 1 TRUE FALSE FALSE FALSE FALSE
E8 NA standard 0.627 -0.007 1 TRUE FALSE FALSE FALSE FALSE
E9 NA standard 0.617 -0.005 1 TRUE FALSE FALSE FALSE FALSE
F1 NA standard 1.107 -0.029 1 TRUE FALSE FALSE FALSE FALSE
F10 NA background 0.156 0.119 1 FALSE FALSE FALSE FALSE TRUE
F11 NA background 0.152 0.119 1 FALSE FALSE FALSE FALSE TRUE
F12 NA background 0.118 0.067 1 FALSE FALSE FALSE FALSE TRUE
F2 NA standard 1.118 -0.018 1 TRUE FALSE FALSE FALSE FALSE
F3 NA standard 1.098 -0.018 1 TRUE FALSE FALSE FALSE FALSE
F4 NA standard 1.370 -0.027 1 TRUE FALSE FALSE FALSE FALSE
F5 NA standard 1.367 -0.024 1 TRUE FALSE FALSE FALSE FALSE
F6 NA standard 1.360 -0.027 1 TRUE FALSE FALSE FALSE FALSE

Wells with at least one anomaly flag

Total flagged wells: 19 / 72 

--- anomaly_table: all flagged wells with their specific flags ---

tibble [19 × 11] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:19] "A7" "E10" "E11" "E12" ...
 $ sample_id         : chr [1:19] "F07_03_ambient" NA NA NA ...
 $ well_type         : chr [1:19] "sample" "standard" "standard" "standard" ...
 $ od_first          : num [1:19] 0.099 0.849 0.854 0.856 0.364 0.373 0.373 0.636 0.627 0.617 ...
 $ net_change        : num [1:19] 0 -0.013 -0.011 -0.013 -0.004 ...
 $ n_flags           : int [1:19] 1 1 1 1 1 1 1 1 1 1 ...
 $ flag_decreasing   : logi [1:19] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ flag_over_range   : logi [1:19] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:19] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:19] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:19] FALSE FALSE FALSE FALSE FALSE FALSE ...

19 of the 72 wells are flagged, in three distinct groups:

  1. 15 GSH standard wells are flagged flag_decreasing – all five non-zero concentrations (8, 16, 24, 32, 40 nmol) end at or below their own starting A412 over the 20-minute window, consistent with TNB2- signal decay (see GSH STANDARD CURVE below). Because the standard curve is calibrated at t = 0 (not from the endpoint), this does not compromise the calibration (R2 = 0.9992, no outlier wells).
  2. All 3 pooled-background wells (F10, F11, F12) are flagged flag_bg_active – the confirmed Reaction Mix mix-up (note 4). See BACKGROUND CORRECTION for what this means for this plate’s activity values.
  3. One sample well, A7 (F07_03_ambient), is flagged flag_decreasing – it is completely flat (net change = 0.000, A412 = 0.099 for all 11 reads), not a slow-rising reaction. Its two sibling replicates (A8, A9) rise normally, so F07_03_ambient’s activity is still computed from 2 of 3 replicates (see RATE EXTRACTION and TECHNICAL REPLICATE PRECISION).

No sample well shows a read glitch or an elevated starting baseline, and no sample well besides A7 fails to rise.

4.3 Background wells behaving as active reactions

bg_active_wells <- well_diagnostics %>%
  filter(well_type == "background", flag_bg_active) %>%
  select(well, sample_id, od_first, net_change)

kable(bg_active_wells, digits = 3,
      caption = "Background wells with a non-flat trace")

cat("Background wells behaving as active reactions:", nrow(bg_active_wells),
    "/", sum(well_diagnostics$well_type == "background"), "\n")

cat("\n--- bg_active_wells: background wells excluded from the background-rate estimate ---\n\n")
str(bg_active_wells)
well sample_id od_first net_change
F10 NA 0.156 0.119
F11 NA 0.152 0.119
F12 NA 0.118 0.067

Background wells with a non-flat trace

Background wells behaving as active reactions: 3 / 3 

--- bg_active_wells: background wells excluded from the background-rate estimate ---

tibble [3 × 4] (S3: tbl_df/tbl/data.frame)
 $ well      : chr [1:3] "F10" "F11" "F12"
 $ sample_id : chr [1:3] NA NA NA
 $ od_first  : num [1:3] 0.156 0.152 0.118
 $ net_change: num [1:3] 0.119 0.119 0.067

All three pooled-background replicates are flagged. F10 and F11 start slightly above the 0.15 A412 threshold (0.156, 0.152) and F12 starts just under it (0.118), but all three fail on drift regardless: net change 0.067-0.119 A412 over 20 minutes, 1.3-2.4x the 0.05 flat-well threshold and of the same order of magnitude as a real sample’s signal. This is the confirmed Reaction Mix mix-up (note 4), not genuine background activity. Because 0 of 3 replicates are flat, none are used in a background-rate estimate – see BACKGROUND CORRECTION.

4.4 Baseline diagnostic (elevated starting absorbance)

bd <- compute_baseline_diagnostics(well_diagnostics, assay_params)
baseline_check       <- bd$baseline_check
baseline_per_sample  <- bd$baseline_per_sample

cat("Sample/background wells with elevated starting A412 (> ",
    assay_params$sample_baseline_od, "):",
    sum(baseline_check$baseline == "elevated"), "/", nrow(baseline_check), "\n", sep = "")

kable(baseline_per_sample, digits = 3,
      caption = "Elevated-baseline replicate count per sample")

cat("\n--- baseline_check: per-well baseline classification ---\n\n")
str(baseline_check)
cat("\n--- baseline_per_sample: elevated-replicate count per sample ---\n\n")
str(baseline_per_sample)
Sample/background wells with elevated starting A412 (> 0.35):0/51
sample_id n_elevated n median_baseline
F07_01_36C 0 3 0.131
F07_01_ambient 0 3 0.115
F07_02_36C 0 3 0.150
F07_02_ambient 0 3 0.147
F07_03_36C 0 3 0.156
F07_03_ambient 0 3 0.140
F07_04_36C 0 3 0.159
F07_04_ambient 0 3 0.126
F07_05_36C 0 3 0.123
F07_05_ambient 0 3 0.141
F07_06_36C 0 3 0.120
F07_06_ambient 0 3 0.151
F07_07_36C 0 3 0.141
F07_07_ambient 0 3 0.159
F07_08_36C 0 3 0.117
F07_08_ambient 0 3 0.128

Elevated-baseline replicate count per sample

--- baseline_check: per-well baseline classification ---

tibble [51 × 25] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:51] "B9" "C10" "B7" "B8" ...
 $ plate_row         : chr [1:51] "B" "C" "B" "B" ...
 $ plate_col         : int [1:51] 9 10 7 8 12 11 8 7 10 5 ...
 $ well_type         : chr [1:51] "sample" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:51] "F07_07_ambient" "F07_04_36C" "F07_07_ambient" "F07_07_ambient" ...
 $ label             : chr [1:51] "F07_07_ambient-citrate_synthase-20.6-df.0" "F07_04_36C-citrate_synthase-16.9-df.0" "F07_07_ambient-citrate_synthase-20.6-df.0" "F07_07_ambient-citrate_synthase-20.6-df.0" ...
 $ std_nmol          : num [1:51] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:51] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:51] 0.161 0.161 0.159 0.159 0.159 0.158 0.157 0.156 0.156 0.153 ...
 $ od_last           : num [1:51] 0.281 0.281 0.277 0.279 0.283 0.279 0.282 0.28 0.275 0.274 ...
 $ od_max            : num [1:51] 0.281 0.281 0.277 0.279 0.283 0.279 0.282 0.28 0.275 0.274 ...
 $ net_change        : num [1:51] 0.12 0.12 0.118 0.12 0.124 0.121 0.125 0.124 0.119 0.121 ...
 $ frac_rising       : num [1:51] 1 1 1 1 1 1 1 1 1 1 ...
 $ max_step          : num [1:51] 0.018 0.018 0.018 0.018 0.019 ...
 $ typical_step      : num [1:51] 0.012 0.012 0.0115 0.0115 0.0125 0.012 0.012 0.0125 0.011 0.0125 ...
 $ step_excess       : num [1:51] -0.001 0.001 0.0015 -0.0015 -0.001 ...
 $ glitch_at_min     : num [1:51] 20 4 14 16 20 16 8 20 4 2 ...
 $ step_ratio        : num [1:51] 1.5 1.5 1.57 1.57 1.52 ...
 $ flag_decreasing   : logi [1:51] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range   : logi [1:51] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:51] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:51] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:51] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:51] 0 0 0 0 0 0 0 0 1 0 ...
 $ baseline          : chr [1:51] "normal" "normal" "normal" "normal" ...

--- baseline_per_sample: elevated-replicate count per sample ---

tibble [16 × 4] (S3: tbl_df/tbl/data.frame)
 $ sample_id      : chr [1:16] "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ n_elevated     : int [1:16] 0 0 0 0 0 0 0 0 0 0 ...
 $ n              : int [1:16] 3 3 3 3 3 3 3 3 3 3 ...
 $ median_baseline: num [1:16] 0.131 0.115 0.15 0.147 0.156 0.14 0.159 0.126 0.123 0.141 ...

No sample or background well anywhere on this plate starts above A412 0.35 — sample wells start within a 0.099-0.161 range (the low end is A7‘s dead well) and the pooled background wells start within 0.118-0.156, both well below the 0.35 threshold. No sample has any elevated-baseline replicate, and none is disqualified on this basis – the background wells’ problem is their drift, not their starting absorbance (note 4).

4.5 Plot starting absorbance

baseline_plot <- ggplot(baseline_check, aes(x = reorder(well, od_first), y = od_first,
                                            fill = well_type)) +
  geom_col() +
  geom_hline(yintercept = assay_params$sample_baseline_od,
             linetype = "dashed", colour = "#cb181d") +
  scale_fill_manual(values = c(sample = "#08519c", background = "#a63603")) +
  labs(title = "Starting A412 (t = 0) for every sample/background well",
       subtitle = paste0("Dashed line = elevated-baseline threshold (A412 = ",
                         assay_params$sample_baseline_od, "). No well on this plate exceeds it."),
       x = "Well", y = "A412 at t = 0", fill = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 90, size = 5, vjust = 0.5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "starting_absorbance.png"), baseline_plot,
       width = 9, height = 6, dpi = 300)

cat("--- baseline_plot: ggplot object structure ---\n\n")
summary(baseline_plot)

baseline_plot

--- baseline_plot: ggplot object structure ---

data: well, plate_row, plate_col, well_type, sample_id, label,
  std_nmol, source, od_first, od_last, od_max, net_change, frac_rising,
  max_step, typical_step, step_excess, glitch_at_min, step_ratio,
  flag_decreasing, flag_over_range, flag_glitch, flag_high_baseline,
  flag_bg_active, n_flags, baseline [51x25]
mapping:  x = ~reorder(well, od_first), y = ~od_first, fill = ~well_type
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

5 GSH STANDARD CURVE

The GSH standard is a stoichiometric endpoint, not an enzymatic reaction: free thiol reduces DTNB immediately on mixing. The standard curve is therefore read from a single timepoint, and t = 0 is used because TNB2- degrades over the run (quantified below).

5.1 Fit the standard curve

Three candidate fits are compared:

  • all wells — every standard replicate, no exclusions
  • concentration means — the conventional fit, on per-concentration means
  • outlier-excluded — replicate-level fit after dropping flagged wells

The outlier-excluded fit is used for activity calculation.

std_curve <- fit_standard_curve(plate_long, assay_params)

standards_t0       <- std_curve$standards_t0
standard_drift     <- std_curve$standard_drift
standard_summary   <- std_curve$standard_summary
standards_flagged  <- std_curve$standards_flagged
fit_all_wells      <- std_curve$fit_all_wells
fit_comparison     <- std_curve$fit_comparison
std_slope          <- std_curve$std_slope
std_intercept      <- std_curve$std_intercept
std_r2             <- std_curve$std_r2
std_nmol_max       <- std_curve$std_nmol_max

cat("Standard wells:", nrow(standards_t0), "at", dplyr::n_distinct(standards_t0$std_nmol),
    "concentrations\n")
cat("Standard drift over 20 min (dA412): median",
    sprintf("%.3f", median(standard_drift$drift)),
    "| range", paste(sprintf("%.3f", range(standard_drift$drift)), collapse = " to "), "\n")
cat("Standards that LOST signal (drift < 0):", sum(standard_drift$drift < 0), "/", nrow(standard_drift), "\n")

cat("\n--- standard_summary: per-concentration mean/SD/CV of t0 absorbance ---\n\n")
str(standard_summary)
kable(standard_summary, digits = c(0, 0, 3, 4, 4, 2, 3, 3, 4),
      caption = "GSH standard curve, per-concentration summary at t = 0")

cat("\nConcentrations exceeding", assay_params$cv_threshold_pct, "% replicate CV:",
    paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
          collapse = ", "), "\n")

cat("\n--- standards_flagged: every standard well with its deviation from the triplicate median ---\n\n")
str(standards_flagged)
kable(standards_flagged %>% filter(is_outlier) %>% select(well, std_nmol, od, triplicate_median, deviation),
      digits = 3, caption = "Standard wells flagged as outliers (|deviation| > 0.15 A412)")

cat("\n--- fit_comparison: slope/intercept/R^2 for all three candidate fits ---\n\n")
str(fit_comparison)
# row.names = FALSE: coef() attaches the predictor's name ("std_nmol") to the
# slope/intercept elements, which data.frame() otherwise promotes to a mangled
# row label (e.g. "std_nmol...1") that breaks pandoc's pipe-table parser.
kable(fit_comparison, digits = c(0, 0, 5, 4, 5), row.names = FALSE,
      caption = "Standard curve fit comparison (outlier-excluded fit used downstream)")
Standard wells: 18 at 6 concentrations
Standard drift over 20 min (dA412): median -0.009 | range -0.029 to 0.002 
Standards that LOST signal (drift < 0): 15 / 18 

--- standard_summary: per-concentration mean/SD/CV of t0 absorbance ---

tibble [6 × 9] (S3: tbl_df/tbl/data.frame)
 $ std_nmol   : num [1:6] 0 8 16 24 32 40
 $ n          : int [1:6] 3 3 3 3 3 3
 $ mean_od    : num [1:6] 0.0953 0.37 0.6267 0.853 1.1077 ...
 $ sd_od      : num [1:6] 0.00462 0.0052 0.0095 0.00361 0.01002 ...
 $ se_od      : num [1:6] 0.00267 0.003 0.00549 0.00208 0.00578 ...
 $ cv_pct     : num [1:6] 4.845 1.404 1.517 0.423 0.904 ...
 $ median_od  : num [1:6] 0.098 0.373 0.627 0.854 1.107 ...
 $ net_od     : num [1:6] 0 0.275 0.531 0.758 1.012 ...
 $ od_per_nmol: num [1:6] NA 0.0343 0.0332 0.0316 0.0316 ...
std_nmol n mean_od sd_od se_od cv_pct median_od net_od od_per_nmol
0 3 0.095 0.0046 0.0027 4.84 0.098 0.000 NA
8 3 0.370 0.0052 0.0030 1.40 0.373 0.275 0.0343
16 3 0.627 0.0095 0.0055 1.52 0.627 0.531 0.0332
24 3 0.853 0.0036 0.0021 0.42 0.854 0.758 0.0316
32 3 1.108 0.0100 0.0058 0.90 1.107 1.012 0.0316
40 3 1.366 0.0051 0.0030 0.38 1.367 1.270 0.0318

GSH standard curve, per-concentration summary at t = 0

Concentrations exceeding 15 % replicate CV:  

--- standards_flagged: every standard well with its deviation from the triplicate median ---

tibble [18 × 7] (S3: tbl_df/tbl/data.frame)
 $ well             : chr [1:18] "F2" "E9" "E4" "E7" ...
 $ std_nmol         : num [1:18] 32 16 8 16 32 0 40 24 40 24 ...
 $ od               : num [1:18] 1.118 0.617 0.364 0.636 1.098 ...
 $ source           : chr [1:18] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ triplicate_median: num [1:18] 1.107 0.627 0.373 0.627 1.107 ...
 $ deviation        : num [1:18] 0.011 -0.01 -0.009 0.009 -0.009 ...
 $ is_outlier       : logi [1:18] FALSE FALSE FALSE FALSE FALSE FALSE ...
well std_nmol od triplicate_median deviation —— ———- —- ——————- ———–

Standard wells flagged as outliers (|deviation| > 0.15 A412)

— fit_comparison: slope/intercept/R^2 for all three candidate fits —

‘data.frame’: 3 obs. of 5 variables: $ fit : chr “all wells” “concentration means” “outlier-excluded” $ n : int 18 6 18 $ slope : num 0.0314 0.0314 0.0314 $ intercept: num 0.108 0.108 0.108 $ r_squared: num 0.999 0.999 0.999

fit n slope intercept r_squared
all wells 18 0.0314 0.1085 0.99925
concentration means 6 0.0314 0.1085 0.99942
outlier-excluded 18 0.0314 0.1085 0.99925

Standard curve fit comparison (outlier-excluded fit used downstream)

No standard well is flagged as an outlier on this plate – every concentration’s triplicate CV is under 5% (highest: the 0 nmol blank at 4.8%; every non-zero concentration is under 1.6%), and no well deviates from its triplicate median by more than the 0.15 A412 threshold. Because there are no outliers to exclude, the “outlier-excluded” and “all wells” fits are identical: slope = 0.03140 A412/nmol, R2 = 0.9992. All 15 non-zero-concentration standard replicates lose signal (drift < 0) between t = 0 and t = 20 min, consistent with TNB2- instability – the 0 nmol blank replicates, which have no thiol to lose, instead drift very slightly upward. This exactly matches the 15 standard wells flagged flag_decreasing above. Because the curve is read at t = 0, this decay affects none of the values used for calibration.

5.2 Plot the standard curve

std_plot <- ggplot(standards_flagged, aes(x = std_nmol, y = od)) +
  geom_point(aes(colour = is_outlier), size = 3) +
  geom_smooth(data = standards_flagged %>% filter(!is_outlier),
              method = "lm", se = TRUE, colour = "#238b45", fill = "#a1d99b") +
  scale_colour_manual(values = c(`TRUE` = "#cb181d", `FALSE` = "#08519c"),
                      labels = c(`TRUE` = "outlier", `FALSE` = "included"),
                      name = "Standard well") +
  labs(title = "GSH standard curve (t = 0), family F07 (ambient and 36°C)",
       subtitle = sprintf("Outlier-excluded fit: slope = %.5f A412/nmol, R² = %.4f (n = %d)",
                          std_slope, std_r2, nrow(standards_flagged %>% filter(!is_outlier))),
       x = "GSH (nmol/well)", y = "A412 at t = 0") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "gsh_standard_curve.png"), std_plot,
       width = 9, height = 6, dpi = 300)

cat("--- std_plot: ggplot object structure ---\n\n")
summary(std_plot)

std_plot

--- std_plot: ggplot object structure ---

data: well, std_nmol, od, source, triplicate_median, deviation,
  is_outlier [18x7]
mapping:  x = ~std_nmol, y = ~od
scales:   colour 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~is_outlier 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

geom_smooth: na.rm = FALSE, orientation = NA, se = TRUE
stat_smooth: na.rm = FALSE, orientation = NA, se = TRUE, method = lm
position_identity

The top standard (40 nmol) stays comfortably within the photometric linear range here – the highest reading is A412 1.37, below the 1.5 ceiling – so the calibration does not need to be restricted to a reduced concentration range.

6 RATE EXTRACTION

Rates are extracted with a sliding-window approach: for each well, the 5-point (8-minute) window of steepest positive slope is taken as the rate, with its R2 retained for the usability check.

well_rates <- compute_well_rates(plate_long, well_diagnostics, assay_params)

cat("--- well_rates: extracted rate window per well ---\n\n")
str(well_rates)
--- well_rates: extracted rate window per well ---

tibble [72 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:72] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:72] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:72] "F07_01_ambient" "F07_04_ambient" "F07_04_ambient" "F07_04_ambient" ...
 $ std_nmol              : num [1:72] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:72] 0 2 2 0 8 2 0 0 0 0 ...
 $ t_end                 : num [1:72] 8 10 10 8 16 10 8 8 8 8 ...
 $ slope_mOD_min         : num [1:72] 3.4 4.9 5 5 3.4 ...
 $ r2                    : num [1:72] 0.999 1 1 1 0.999 ...
 $ max_abs_slope_mOD_min : num [1:72] 3.4 4.9 5 5 3.4 ...
 $ abs_window_is_negative: logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:72] 0.066 0.094 0.094 0.095 0.067 0.066 0.116 0.12 0.117 0 ...
 $ frac_rising           : num [1:72] 1 1 1 1 1 1 1 1 1 0 ...
 $ od_max                : num [1:72] 0.18 0.22 0.22 0.221 0.182 0.181 0.263 0.268 0.263 0.099 ...
 $ flag_decreasing       : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:72] 20 14 18 10 18 2 2 6 2 2 ...
 $ flag_bg_active        : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:72] 0 0 0 0 0 0 0 0 0 1 ...
 $ glitch_in_window      : logi [1:72] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:72] TRUE TRUE TRUE TRUE TRUE TRUE ...

6.1 Rate tables by well type

sample_rates <- well_rates %>% filter(well_type == "sample")

kable(sample_rates %>% select(well, sample_id, t_start, t_end, slope_mOD_min, r2, rate_usable),
      digits = c(0, 0, 0, 0, 3, 4, 0),
      caption = "Extracted rate window, sample wells")

cat("\nSample wells with a usable rate:", sum(sample_rates$rate_usable), "/", nrow(sample_rates), "\n")
cat("Sample rate range (mA412/min):",
    paste(sprintf("%.2f", range(sample_rates$slope_mOD_min)), collapse = " - "), "\n")
cat("Median replicate-level R^2:", sprintf("%.4f", median(sample_rates$r2)), "\n")

cat("\n--- sample_rates: rate-window results restricted to sample wells ---\n\n")
str(sample_rates)
well sample_id t_start t_end slope_mOD_min r2 rate_usable
A1 F07_01_ambient 0 8 3.40 0.9991 TRUE
A10 F07_04_ambient 2 10 4.90 0.9996 TRUE
A11 F07_04_ambient 2 10 5.00 1.0000 TRUE
A12 F07_04_ambient 0 8 5.00 1.0000 TRUE
A2 F07_01_ambient 8 16 3.40 0.9991 TRUE
A3 F07_01_ambient 2 10 3.50 1.0000 TRUE
A4 F07_02_ambient 0 8 7.35 0.9991 TRUE
A5 F07_02_ambient 0 8 7.60 0.9993 TRUE
A6 F07_02_ambient 0 8 7.35 0.9991 TRUE
A7 F07_03_ambient 0 8 0.00 0.0000 FALSE
A8 F07_03_ambient 0 8 6.85 0.9990 TRUE
A9 F07_03_ambient 0 8 7.50 0.9995 TRUE
B1 F07_05_ambient 0 8 6.50 1.0000 TRUE
B10 F07_08_ambient 2 10 4.90 0.9996 TRUE
B11 F07_08_ambient 0 8 4.85 0.9997 TRUE
B12 F07_08_ambient 6 14 4.90 0.9996 TRUE
B2 F07_05_ambient 0 8 6.65 0.9989 TRUE
B3 F07_05_ambient 0 8 6.45 0.9993 TRUE
B4 F07_06_ambient 0 8 7.10 0.9986 TRUE
B5 F07_06_ambient 0 8 7.35 0.9991 TRUE
B6 F07_06_ambient 0 8 7.15 0.9995 TRUE
B7 F07_07_ambient 0 8 8.15 0.9978 TRUE
B8 F07_07_ambient 0 8 8.25 0.9987 TRUE
B9 F07_07_ambient 0 8 8.25 0.9987 TRUE
C1 F07_01_36C 4 12 4.90 0.9996 TRUE
C10 F07_04_36C 0 8 8.40 0.9983 TRUE
C11 F07_04_36C 0 8 8.50 0.9975 TRUE
C12 F07_04_36C 0 8 8.65 0.9980 TRUE
C2 F07_01_36C 4 12 5.00 0.9992 TRUE
C3 F07_01_36C 4 12 4.90 0.9996 TRUE
C4 F07_02_36C 0 8 7.35 0.9991 TRUE
C5 F07_02_36C 0 8 7.50 0.9995 TRUE
C6 F07_02_36C 0 8 7.60 0.9998 TRUE
C7 F07_03_36C 0 8 8.35 0.9993 TRUE
C8 F07_03_36C 0 8 8.50 0.9990 TRUE
C9 F07_03_36C 0 8 8.00 0.9995 TRUE
D1 F07_05_36C 2 10 4.40 0.9995 TRUE
D10 F07_08_36C 4 12 3.65 0.9994 TRUE
D11 F07_08_36C 6 14 3.65 0.9994 TRUE
D12 F07_08_36C 4 12 3.65 0.9994 TRUE
D2 F07_05_36C 2 10 4.35 0.9996 TRUE
D3 F07_05_36C 2 10 4.35 0.9996 TRUE
D4 F07_06_36C 6 14 3.65 0.9994 TRUE
D5 F07_06_36C 4 12 3.80 0.9993 TRUE
D6 F07_06_36C 4 12 3.65 0.9994 TRUE
D7 F07_07_36C 0 8 7.10 0.9998 TRUE
D8 F07_07_36C 0 8 7.00 1.0000 TRUE
D9 F07_07_36C 0 8 7.05 0.9994 TRUE

Extracted rate window, sample wells

Sample wells with a usable rate: 47 / 48 
Sample rate range (mA412/min): 0.00 - 8.65 
Median replicate-level R^2: 0.9994 

--- sample_rates: rate-window results restricted to sample wells ---

tibble [48 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:48] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:48] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:48] "F07_01_ambient" "F07_04_ambient" "F07_04_ambient" "F07_04_ambient" ...
 $ std_nmol              : num [1:48] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:48] 0 2 2 0 8 2 0 0 0 0 ...
 $ t_end                 : num [1:48] 8 10 10 8 16 10 8 8 8 8 ...
 $ slope_mOD_min         : num [1:48] 3.4 4.9 5 5 3.4 ...
 $ r2                    : num [1:48] 0.999 1 1 1 0.999 ...
 $ max_abs_slope_mOD_min : num [1:48] 3.4 4.9 5 5 3.4 ...
 $ abs_window_is_negative: logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:48] 0.066 0.094 0.094 0.095 0.067 0.066 0.116 0.12 0.117 0 ...
 $ frac_rising           : num [1:48] 1 1 1 1 1 1 1 1 1 0 ...
 $ od_max                : num [1:48] 0.18 0.22 0.22 0.221 0.182 0.181 0.263 0.268 0.263 0.099 ...
 $ flag_decreasing       : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:48] 20 14 18 10 18 2 2 6 2 2 ...
 $ flag_bg_active        : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:48] 0 0 0 0 0 0 0 0 0 1 ...
 $ glitch_in_window      : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:48] TRUE TRUE TRUE TRUE TRUE TRUE ...

47 of the 48 sample wells clear the R2 >= 0.80 and positive-net-change usability bar; every usable replicate-level R2 exceeds 0.997 (median 0.9994). The one exclusion is A7 (F07_03_ambient), the dead well noted above (slope = 0.00 mA412/min, R2 = 0, rate_usable = FALSE); its two siblings (A8, A9) are both usable. Among the 47 usable wells, rates range 3.40-8.65 mA412/min.

6.2 Positive control

pos_control <- compute_positive_control(well_rates, std_curve$std_slope, assay_params)

kable(pos_control %>% select(well, t_start, t_end, slope_mOD_min, r2, net_change, activity_mU_uL),
      digits = c(0, 0, 0, 3, 4, 3, 4),
      caption = "Positive control replicates")

cat("Positive control mean rate:", sprintf("%.3f", mean(pos_control$slope_mOD_min)), "mA412/min\n")
cat("Positive control CV:", sprintf("%.1f",
    100 * sd(pos_control$slope_mOD_min) / mean(pos_control$slope_mOD_min)), "%\n")
cat("All replicates rising and linear (R^2 > 0.99):",
    all(pos_control$net_change > 0 & pos_control$r2 > 0.99), "\n")

cat("\n--- pos_control: positive control rate and R^2 per replicate ---\n\n")
str(pos_control)
well t_start t_end slope_mOD_min r2 net_change activity_mU_uL
F7 12 20 2.95 0.9980 0.055 0.0470
F8 8 16 2.85 0.9991 0.054 0.0454
F9 12 20 3.00 1.0000 0.055 0.0478

Positive control replicates

Positive control mean rate: 2.933 mA412/min
Positive control CV: 2.6 %
All replicates rising and linear (R^2 > 0.99): TRUE 

--- pos_control: positive control rate and R^2 per replicate ---

tibble [3 × 22] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:3] "F7" "F8" "F9"
 $ well_type             : chr [1:3] "positive_control" "positive_control" "positive_control"
 $ sample_id             : chr [1:3] NA NA NA
 $ std_nmol              : num [1:3] NA NA NA
 $ t_start               : num [1:3] 12 8 12
 $ t_end                 : num [1:3] 20 16 20
 $ slope_mOD_min         : num [1:3] 2.95 2.85 3
 $ r2                    : num [1:3] 0.998 0.999 1
 $ max_abs_slope_mOD_min : num [1:3] 2.95 2.85 3
 $ abs_window_is_negative: logi [1:3] FALSE FALSE FALSE
 $ net_change            : num [1:3] 0.055 0.054 0.055
 $ frac_rising           : num [1:3] 1 1 1
 $ od_max                : num [1:3] 0.295 0.298 0.297
 $ flag_decreasing       : logi [1:3] FALSE FALSE FALSE
 $ flag_over_range       : logi [1:3] FALSE FALSE FALSE
 $ flag_glitch           : logi [1:3] FALSE FALSE FALSE
 $ glitch_at_min         : num [1:3] 20 4 2
 $ flag_bg_active        : logi [1:3] FALSE FALSE FALSE
 $ n_flags               : int [1:3] 0 0 0
 $ glitch_in_window      : logi [1:3] FALSE FALSE FALSE
 $ rate_usable           : logi [1:3] TRUE TRUE TRUE
 $ activity_mU_uL        : num [1:3] 0.047 0.0454 0.0478

The positive control rose linearly in all three replicates (R2 0.998-1.000) at a consistent 2.85-3.00 mA412/min, a 2.6% CV across the triplicate – confirming that the core reaction chemistry and reader were functioning normally on this plate. This is a tight, well-behaved positive control.

7 BACKGROUND CORRECTION

This plate uses a single pooled background control (note 3): the background rate is normally estimated once, from the flat (well-behaved) replicates of the BG-citrate_synthase triplicate, and the same corrected rate is subtracted from every sample. On this plate, none of the three background replicates are flat (note 4) – compute_background_correction() therefore cannot produce a background-rate estimate, and its documented fallback applies: a missing (NA) bg_rate_flat is treated as 0. Every sample’s corrected rate on this plate is therefore identical to its raw rate.

background_correction <- compute_background_correction(well_rates)

cat("Pooled background rate (flat wells only):\n")
print(background_correction, row.names = FALSE, digits = 3)

cat("\nBackground rate used for correction:",
    sprintf("%.3f", background_correction$bg_rate_flat), "mA412/min (from",
    background_correction$n_bg_flat, "of", background_correction$n_bg_total, "replicates)\n")
cat("Median sample rate:",
    sprintf("%.2f", median(sample_rates$slope_mOD_min)), "mA412/min\n")
cat("Background as % of median sample signal:",
    sprintf("%.1f", 100 * background_correction$bg_rate_flat /
                    median(sample_rates$slope_mOD_min)), "%\n")

cat("\n--- background_correction: pooled background rate estimate ---\n\n")
str(background_correction)
Pooled background rate (flat wells only):
 n_bg_total n_bg_flat bg_rate_flat bg_rate_all
          3         0           NA         7.7

Background rate used for correction: NA mA412/min (from 0 of 3 replicates)
Median sample rate: 6.58 mA412/min
Background as % of median sample signal: NA %

--- background_correction: pooled background rate estimate ---

'data.frame':   1 obs. of  4 variables:
 $ n_bg_total  : int 3
 $ n_bg_flat   : int 0
 $ bg_rate_flat: num NA
 $ bg_rate_all : num 7.7

n_bg_flat is 0 of 3, so bg_rate_flat is NA and is treated as 0 downstream. For reference only (not used in any activity calculation), the mean rate across all three – compromised – background replicates is 7.7 mA412/min, actually higher than the median sample rate of 6.58 mA412/min, underscoring that these wells behaved as active reactions rather than anything resembling a background control.

7.1 Background significance test

The flat/active threshold classification above is backed by an explicit significance test: a Welch two-sample t-test of well-level rate (background vs. sample), plus an ANOVA on the OD trajectory over time. This mirrors the one-time background ANOVA validation done by Cattau et al. (2023) on the legacy version of this assay. On a well-behaved plate this test should show background wells statistically slower than sample wells; on this plate, given the confirmed mix-up, no such difference is expected – it is run here mainly to document that expectation quantitatively rather than as a validity check.

bg_sample_rates <- well_rates %>%
  filter(well_type %in% c("sample", "background"))

bg_significance_test <- t.test(slope_mOD_min ~ well_type, data = bg_sample_rates)

cat("--- Welch two-sample t-test: well-level rate (background vs. sample) ---\n\n")
print(bg_significance_test)

bg_trend_anova <- anova(lm(od ~ well_type * time_min,
                            data = plate_long %>% filter(well_type %in% c("sample", "background"))))

cat("\n--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---\n\n")
print(bg_trend_anova)

cat("\nInterpretation: mean background-well rate is",
    sprintf("%.2f", unname(bg_significance_test$estimate["mean in group background"])),
    "mA412/min vs.", sprintf("%.2f", unname(bg_significance_test$estimate["mean in group sample"])),
    "mA412/min for sample wells (p =", format.pval(bg_significance_test$p.value, digits = 3),
    "). Background wells are NOT statistically slower than sample wells on\n",
    "this plate (n = 3 background replicates limits power, but the point\n",
    "estimate itself is not lower) -- consistent with the confirmed Reaction\n",
    "Mix mix-up (note 4) rather than a genuine, if imprecise, background.\n")

cat("\n--- bg_significance_test: Welch t-test result object ---\n\n")
str(bg_significance_test)

cat("\n--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---\n\n")
str(bg_trend_anova)
--- Welch two-sample t-test: well-level rate (background vs. sample) ---


    Welch Two Sample t-test

data:  slope_mOD_min by well_type
t = 1.3803, df = 2.2186, p-value = 0.2901
alternative hypothesis: true difference in means between group background and group sample is not equal to 0
95 percent confidence interval:
 -3.194340  6.665174
sample estimates:
mean in group background     mean in group sample 
                7.700000                 5.964583 


--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---

Analysis of Variance Table

Response: od
                    Df  Sum Sq Mean Sq  F value  Pr(>F)    
well_type            1 0.00494 0.00494   4.5694 0.03298 *  
time_min             1 0.57624 0.57624 532.8372 < 2e-16 ***
well_type:time_min   1 0.00004 0.00004   0.0332 0.85558    
Residuals          557 0.60237 0.00108                     
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Interpretation: mean background-well rate is 7.70 mA412/min vs. 5.96 mA412/min for sample wells (p = 0.29 ). Background wells are NOT statistically slower than sample wells on
 this plate (n = 3 background replicates limits power, but the point
 estimate itself is not lower) -- consistent with the confirmed Reaction
 Mix mix-up (note 4) rather than a genuine, if imprecise, background.

--- bg_significance_test: Welch t-test result object ---

List of 10
 $ statistic  : Named num 1.38
  ..- attr(*, "names")= chr "t"
 $ parameter  : Named num 2.22
  ..- attr(*, "names")= chr "df"
 $ p.value    : num 0.29
 $ conf.int   : num [1:2] -3.19 6.67
  ..- attr(*, "conf.level")= num 0.95
 $ estimate   : Named num [1:2] 7.7 5.96
  ..- attr(*, "names")= chr [1:2] "mean in group background" "mean in group sample"
 $ null.value : Named num 0
  ..- attr(*, "names")= chr "difference in means between group background and group sample"
 $ stderr     : num 1.26
 $ alternative: chr "two.sided"
 $ method     : chr "Welch Two Sample t-test"
 $ data.name  : chr "slope_mOD_min by well_type"
 - attr(*, "class")= chr "htest"

--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---

Classes 'anova' and 'data.frame':   4 obs. of  5 variables:
 $ Df     : int  1 1 1 557
 $ Sum Sq : num  4.94e-03 5.76e-01 3.59e-05 6.02e-01
 $ Mean Sq: num  4.94e-03 5.76e-01 3.59e-05 1.08e-03
 $ F value: num  4.5694 532.8372 0.0332 NA
 $ Pr(>F) : num  3.30e-02 3.15e-83 8.56e-01 NA
 - attr(*, "heading")= chr [1:2] "Analysis of Variance Table\n" "Response: od"

8 TECHNICAL REPLICATE PRECISION

Coefficient of variation across the three technical replicates of each sample, computed on the extracted rate. Both all and usable replicates only are reported; they differ only for F07_03_ambient, whose dead A7 well is excluded from the “usable” calculation.

cv_summary <- compute_replicate_cv(well_rates, protein_by_sample$sample_id, assay_params)

kable(cv_summary %>% select(sample_id, n_all, mean_all, sd_all, cv_all,
                            n_usable, mean_usable, sd_usable, cv_usable,
                            excluded_wells),
      digits = c(0, 0, 2, 2, 1, 0, 2, 2, 1, 0),
      col.names = c("Sample", "n", "Mean rate", "SD", "CV (%)",
                    "n usable", "Mean rate", "SD", "CV (%)", "Excluded wells"),
      caption = paste0("Technical replicate CV of the extracted rate (mA412/min), ",
                       "all replicates vs. usable replicates only"))

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on ALL replicates:", sum(cv_summary$fails_cv_all), "/", nrow(cv_summary), "\n")
print(cv_summary %>% filter(fails_cv_all) %>%
        select(sample_id, n_all, mean_all, sd_all, cv_all) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on USABLE replicates:", sum(cv_summary$fails_cv_usable), "\n")
print(cv_summary %>% filter(fails_cv_usable) %>%
        select(sample_id, n_usable, mean_usable, sd_usable, cv_usable) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

write.csv(cv_summary, file.path(output_dir, "technical_replicate_cv.csv"), row.names = FALSE)

cat("\n--- cv_summary: per-sample technical replicate CV, both variants ---\n\n")
str(cv_summary)
Sample n Mean rate SD CV (%) n usable Mean rate SD CV (%) Excluded wells
F07_01_36C 3 4.93 0.06 1.2 3 4.93 0.06 1.2 -
F07_01_ambient 3 3.43 0.06 1.7 3 3.43 0.06 1.7 -
F07_02_36C 3 7.48 0.13 1.7 3 7.48 0.13 1.7 -
F07_02_ambient 3 7.43 0.14 1.9 3 7.43 0.14 1.9 -
F07_03_36C 3 8.28 0.26 3.1 3 8.28 0.26 3.1 -
F07_03_ambient 3 4.78 4.16 86.9 2 7.18 0.46 6.4 A7
F07_04_36C 3 8.52 0.13 1.5 3 8.52 0.13 1.5 -
F07_04_ambient 3 4.97 0.06 1.2 3 4.97 0.06 1.2 -
F07_05_36C 3 4.37 0.03 0.7 3 4.37 0.03 0.7 -
F07_05_ambient 3 6.53 0.10 1.6 3 6.53 0.10 1.6 -
F07_06_36C 3 3.70 0.09 2.3 3 3.70 0.09 2.3 -
F07_06_ambient 3 7.20 0.13 1.8 3 7.20 0.13 1.8 -
F07_07_36C 3 7.05 0.05 0.7 3 7.05 0.05 0.7 -
F07_07_ambient 3 8.22 0.06 0.7 3 8.22 0.06 0.7 -
F07_08_36C 3 3.65 0.00 0.0 3 3.65 0.00 0.0 -
F07_08_ambient 3 4.88 0.03 0.6 3 4.88 0.03 0.6 -

Technical replicate CV of the extracted rate (mA412/min), all replicates vs. usable replicates only

Samples exceeding CV 15% on ALL replicates: 1 / 16 
      sample_id n_all mean_all sd_all cv_all
 F07_03_ambient     3    4.783  4.155  86.87

Samples exceeding CV 15% on USABLE replicates: 0 
[1] sample_id   n_usable    mean_usable sd_usable   cv_usable  
<0 rows> (or 0-length row.names)

--- cv_summary: per-sample technical replicate CV, both variants ---

'data.frame':   16 obs. of  12 variables:
 $ sample_id      : chr  "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ n_all          : int  3 3 3 3 3 3 3 3 3 3 ...
 $ mean_all       : num  4.93 3.43 7.48 7.43 8.28 ...
 $ sd_all         : num  0.0577 0.0577 0.1258 0.1443 0.2566 ...
 $ cv_all         : num  1.17 1.68 1.68 1.94 3.1 ...
 $ n_usable       : int  3 3 3 3 3 2 3 3 3 3 ...
 $ mean_usable    : num  4.93 3.43 7.48 7.43 8.28 ...
 $ sd_usable      : num  0.0577 0.0577 0.1258 0.1443 0.2566 ...
 $ cv_usable      : num  1.17 1.68 1.68 1.94 3.1 ...
 $ excluded_wells : Named chr  "-" "-" "-" "-" ...
  ..- attr(*, "names")= chr [1:16] "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ fails_cv_all   : logi  FALSE FALSE FALSE FALSE FALSE TRUE ...
 $ fails_cv_usable: logi  FALSE FALSE FALSE FALSE FALSE FALSE ...

Only one sample exceeds the 15% CV threshold, and only on the “all replicates” metric. F07_03_ambient shows an 86.9% CV across all three raw replicates, entirely driven by its dead A7 well (slope 0.00 vs. 6.85 and 7.50 mA412/min for A8/A9); its usable-replicate CV is a normal 6.4%. Every other sample is well under the threshold on both metrics – the highest of the remaining 15 is F07_03_36C at 3.1%, and most are under 2%.

8.1 Plot replicate spread

cv_plot_data <- sample_rates %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id")

cv_plot <- ggplot(cv_plot_data, aes(x = sample_id, y = slope_mOD_min)) +
  geom_point(aes(colour = rate_usable, shape = rate_usable),
             size = 3, position = position_jitter(width = 0.12, height = 0)) +
  stat_summary(fun = mean, geom = "crossbar", width = 0.5,
               colour = "grey30", linewidth = 0.3) +
  geom_text(data = cv_summary, aes(x = sample_id, y = Inf,
                                   label = sprintf("CV %.0f%%", cv_all)),
            vjust = 1.6, size = 3,
            colour = ifelse(cv_summary$fails_cv_all, "#cb181d", "grey30")) +
  scale_colour_manual(values = c(`TRUE` = "#08519c", `FALSE` = "#cb181d"),
                      labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                     labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  labs(title = "Technical replicate spread in extracted rate",
       subtitle = paste0("Crossbar = mean of all three replicates; CV computed on all three. ",
                         "Threshold ", assay_params$cv_threshold_pct, "%."),
       x = "Sample", y = "Rate (mA412/min)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 30, hjust = 1))

ggsave(file.path(output_dir, "technical_replicate_cv.png"), cv_plot,
       width = 9, height = 6, dpi = 300)

cat("--- cv_plot: ggplot object structure ---\n\n")
summary(cv_plot)

cv_plot

--- cv_plot: ggplot object structure ---

data: well, well_type, sample_id, std_nmol, t_start, t_end,
  slope_mOD_min, r2, max_abs_slope_mOD_min, abs_window_is_negative,
  net_change, frac_rising, od_max, flag_decreasing, flag_over_range,
  flag_glitch, glitch_at_min, flag_bg_active, n_flags,
  glitch_in_window, rate_usable, cv_all [48x22]
mapping:  x = ~sample_id, y = ~slope_mOD_min
scales:   colour, shape 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~rate_usable, shape = ~rate_usable 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_jitter 

geom_crossbar: na.rm = FALSE, orientation = NA, width = 0.5
stat_summary: fun.data = NULL, fun = function (x, ...) 
UseMethod("mean"), fun.max = NULL, fun.min = NULL, fun.args = list(), na.rm = FALSE, orientation = NA
position_identity 

mapping: x = ~sample_id, y = Inf, label = ~sprintf("CV %.0f%%", cv_all) 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge

9 CITRATE SYNTHASE ACTIVITY

9.1 Calculation

Following Abcam Section 10.3:

Sample CS activity = B / (dT x V) x D nmol/min/µL (= mU/µL)

  1. Rate (A412/min) = max-increasing sliding-window slope / 1000
  2. Background-corrected rate = sample rate - the single pooled background rate (note 3), applied identically to every sample – on this plate that background rate is 0 for every sample (note 4), so corrected rate = raw rate throughout
  3. B/dT (nmol/min) = background-corrected rate / standard curve slope (0.03140 A412 per nmol, from the outlier-excluded fit)
  4. Activity (mU/µL) = (B/dT) / V x D, with V = 2 µL and D = 1
  5. Protein-normalized (mU/mg protein) = activity x homogenate_volume_uL (350 µL, measured) / total protein (mg)
cs_activity <- calculate_cs_activity(well_rates, protein_by_sample, background_correction,
                                      std_slope, std_nmol_max, assay_params, plate_long)

cat("Samples with a computed activity:", nrow(cs_activity), "/", nrow(protein_by_sample), "\n")
cat("Activity range (mU/uL):",
    paste(sprintf("%.4f", range(cs_activity$activity_mU_per_uL)), collapse = " - "), "\n")
cat("Activity range (mU/mg protein):",
    paste(sprintf("%.3f", range(cs_activity$activity_mU_per_mg_protein)), collapse = " - "), "\n")

cat("\n--- cs_activity: per-sample citrate synthase activity ---\n\n")
str(cs_activity)
Samples with a computed activity: 16 / 16 
Activity range (mU/uL): 0.0547 - 0.1356 
Activity range (mU/mg protein): 79.566 - 207.819 

--- cs_activity: per-sample citrate synthase activity ---

tibble [16 × 19] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:16] "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ n_reps_used               : int [1:16] 3 3 3 3 3 2 3 3 3 3 ...
 $ mean_rate_mOD_min         : num [1:16] 4.93 3.43 7.48 7.43 8.28 ...
 $ sd_rate                   : num [1:16] 0.0577 0.0577 0.1258 0.1443 0.2566 ...
 $ cv_rate                   : num [1:16] 1.17 1.68 1.68 1.94 3.1 ...
 $ family                    : chr [1:16] "F07" "F07" "F07" "F07" ...
 $ individual                : chr [1:16] "01" "01" "02" "02" ...
 $ temperature               : chr [1:16] "36C" "ambient" "36C" "ambient" ...
 $ conc_ug_mL                : num [1:16] 663 263 932 1080 1206 ...
 $ total_protein_mg          : num [1:16] 0.2322 0.0921 0.3263 0.378 0.4219 ...
 $ bg_rate_mOD_min           : num [1:16] 0 0 0 0 0 0 0 0 0 0 ...
 $ corrected_mOD_min         : num [1:16] 4.93 3.43 7.48 7.43 8.28 ...
 $ rate_OD_min               : num [1:16] 0.00493 0.00343 0.00748 0.00743 0.00828 ...
 $ nmol_per_min              : num [1:16] 0.157 0.109 0.238 0.237 0.264 ...
 $ activity_mU_per_uL        : num [1:16] 0.0786 0.0547 0.1192 0.1184 0.1319 ...
 $ total_mU_in_homogenate    : num [1:16] 27.5 19.1 41.7 41.4 46.2 ...
 $ activity_mU_per_mg_protein: num [1:16] 118 208 128 110 109 ...
 $ nmol_in_window            : num [1:16] 3.14 2.19 4.77 4.74 5.28 ...
 $ within_std_range          : logi [1:16] TRUE TRUE TRUE TRUE TRUE TRUE ...

9.2 Results table

results_table <- build_results_table(cs_activity, cv_summary, baseline_per_sample, assay_params)

kable(results_table,
      caption = paste0("Citrate synthase activity, family F07, ambient and 36°C (mU/mg protein uses ",
                       "measured protein concentration x the ", assay_params$homogenate_volume_uL,
                       " uL homogenization volume)"))

cat("\nInterpretation summary:\n")
print(results_table %>% count(Interpretation, name = "n_samples") %>%
        as.data.frame(), row.names = FALSE)

write.csv(results_table, file.path(output_dir, "citrate_synthase_activity_results.csv"),
          row.names = FALSE)

cat("\n--- results_table: formatted per-sample results written to CSV ---\n\n")
str(results_table)
Sample Family Individual Temperature Protein conc (ug/mL) Total protein (mg) Reps used CV all reps (%) CV used reps (%) Rate (mA412/min) BG rate (mA412/min) Corrected rate (mA412/min) Activity (mU/uL) Activity (mU/mg protein) CV flag Elevated baseline reps Interpretation
F07_01_36C F07 01 36C 663.3 0.232 3/3 1.2 1.2 4.93 0 4.93 0.0786 118.446 pass 0/3 usable
F07_01_ambient F07 01 ambient 263.1 0.092 3/3 1.7 1.7 3.43 0 3.43 0.0547 207.819 pass 0/3 usable
F07_02_36C F07 02 36C 932.4 0.326 3/3 1.7 1.7 7.48 0 7.48 0.1192 127.815 pass 0/3 usable
F07_02_ambient F07 02 ambient 1080.1 0.378 3/3 1.9 1.9 7.43 0 7.43 0.1184 109.600 pass 0/3 usable
F07_03_36C F07 03 36C 1205.5 0.422 3/3 3.1 3.1 8.28 0 8.28 0.1319 109.428 pass 0/3 usable
F07_03_ambient F07 03 ambient 1155.6 0.404 2/3 86.9 6.4 7.18 0 7.18 0.1143 98.879 FAIL >15% 0/3 usable
F07_04_36C F07 04 36C 1567.2 0.549 3/3 1.5 1.5 8.52 0 8.52 0.1356 86.544 pass 0/3 usable
F07_04_ambient F07 04 ambient 560.0 0.196 3/3 1.2 1.2 4.97 0 4.97 0.0791 141.243 pass 0/3 usable
F07_05_36C F07 05 36C 427.9 0.150 3/3 0.7 0.7 4.37 0 4.37 0.0695 162.517 pass 0/3 usable
F07_05_ambient F07 05 ambient 659.3 0.231 3/3 1.6 1.6 6.53 0 6.53 0.1040 157.813 pass 0/3 usable
F07_06_36C F07 06 36C 496.5 0.174 3/3 2.3 2.3 3.70 0 3.70 0.0589 118.679 pass 0/3 usable
F07_06_ambient F07 06 ambient 1119.4 0.392 3/3 1.8 1.8 7.20 0 7.20 0.1147 102.432 pass 0/3 usable
F07_07_36C F07 07 36C 1038.1 0.363 3/3 0.7 0.7 7.05 0 7.05 0.1123 108.153 pass 0/3 usable
F07_07_ambient F07 07 ambient 1644.6 0.576 3/3 0.7 0.7 8.22 0 8.22 0.1309 79.566 pass 0/3 usable
F07_08_36C F07 08 36C 289.8 0.101 3/3 0.0 0.0 3.65 0 3.65 0.0581 200.578 pass 0/3 usable
F07_08_ambient F07 08 ambient 560.2 0.196 3/3 0.6 0.6 4.88 0 4.88 0.0778 138.824 pass 0/3 usable

Citrate synthase activity, family F07, ambient and 36°C (mU/mg protein uses measured protein concentration x the 350 uL homogenization volume)

Interpretation summary:
 Interpretation n_samples
         usable        16

--- results_table: formatted per-sample results written to CSV ---

tibble [16 × 17] (S3: tbl_df/tbl/data.frame)
 $ Sample                    : chr [1:16] "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ Family                    : chr [1:16] "F07" "F07" "F07" "F07" ...
 $ Individual                : chr [1:16] "01" "01" "02" "02" ...
 $ Temperature               : chr [1:16] "36C" "ambient" "36C" "ambient" ...
 $ Protein conc (ug/mL)      : num [1:16] 663 263 932 1080 1206 ...
 $ Total protein (mg)        : num [1:16] 0.232 0.092 0.326 0.378 0.422 0.404 0.549 0.196 0.15 0.231 ...
 $ Reps used                 : chr [1:16] "3/3" "3/3" "3/3" "3/3" ...
 $ CV all reps (%)           : num [1:16] 1.2 1.7 1.7 1.9 3.1 86.9 1.5 1.2 0.7 1.6 ...
 $ CV used reps (%)          : num [1:16] 1.2 1.7 1.7 1.9 3.1 6.4 1.5 1.2 0.7 1.6 ...
 $ Rate (mA412/min)          : num [1:16] 4.93 3.43 7.48 7.43 8.28 7.18 8.52 4.97 4.37 6.53 ...
 $ BG rate (mA412/min)       : num [1:16] 0 0 0 0 0 0 0 0 0 0 ...
 $ Corrected rate (mA412/min): num [1:16] 4.93 3.43 7.48 7.43 8.28 7.18 8.52 4.97 4.37 6.53 ...
 $ Activity (mU/uL)          : num [1:16] 0.0786 0.0547 0.1192 0.1184 0.1319 ...
 $ Activity (mU/mg protein)  : num [1:16] 118 208 128 110 109 ...
 $ CV flag                   : chr [1:16] "pass" "pass" "pass" "pass" ...
 $ Elevated baseline reps    : chr [1:16] "0/3" "0/3" "0/3" "0/3" ...
 $ Interpretation            : chr [1:16] "usable" "usable" "usable" "usable" ...

All 16 samples are classified usable in the Interpretation column – calculate_cs_activity() already restricts each sample’s mean rate to its usable replicates, so F07_03_ambient’s activity value is computed from A8/A9 only and its usable-replicate CV (6.4%) is what determines its Interpretation. Its CV flag column still reads FAIL >15%, because that column is deliberately computed on all raw replicates (including the dead A7) as a stricter, unfiltered check – both are correct, and they are reporting different things (see TECHNICAL REPLICATE PRECISION above).

9.3 Plot activity

# Three-level quality scheme. F07_03_ambient falls in the "CV > 15%" tier
# here (cv_all, unfiltered) even though its activity value itself is
# "usable" in the results table above (cv_rate, usable-replicates only).
qual_levels <- c("clean", paste0("CV > ", assay_params$cv_threshold_pct, "%"),
                 "all reps baseline-compromised")
qual_cols   <- setNames(c("#08519c", "#cb181d", "grey60"), qual_levels)

activity_plot_data <- cs_activity %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id") %>%
  left_join(baseline_per_sample %>% select(sample_id, n_elevated, n), by = "sample_id") %>%
  mutate(cv_flag = factor(case_when(
                            n_elevated == n                              ~ qual_levels[3],
                            cv_all > assay_params$cv_threshold_pct       ~ qual_levels[2],
                            TRUE                                        ~ qual_levels[1]),
                          levels = qual_levels),
         label_n = paste0(n_reps_used, "/3 reps",
                          ifelse(n_elevated == n, " - DISQUALIFIED", "")))

activity_plot <- ggplot(activity_plot_data,
                        aes(x = reorder(sample_id, activity_mU_per_uL),
                            y = activity_mU_per_uL, fill = cv_flag)) +
  geom_col(width = 0.65) +
  geom_text(aes(label = label_n), hjust = -0.15, size = 3, colour = "grey25") +
  scale_fill_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  coord_flip(clip = "off") +
  expand_limits(y = max(activity_plot_data$activity_mU_per_uL) * 1.25) +
  labs(title = "Citrate synthase activity, M. gigas ctenidia, family F07 (ambient and 36°C)",
       subtitle = paste0("Not background-corrected (bg = 0 for every sample, note 4); mean of usable ",
                         "technical replicates (2/3 for F07_03_ambient, 3/3 for every other sample)."),
       x = NULL, y = "CS activity (mU/uL homogenate)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "citrate_synthase_activity.png"), activity_plot,
       width = 10, height = 6, dpi = 300)

cat("--- activity_plot: ggplot object structure ---\n\n")
summary(activity_plot)

activity_plot

--- activity_plot: ggplot object structure ---

data: sample_id, n_reps_used, mean_rate_mOD_min, sd_rate, cv_rate,
  family, individual, temperature, conc_ug_mL, total_protein_mg,
  bg_rate_mOD_min, corrected_mOD_min, rate_OD_min, nmol_per_min,
  activity_mU_per_uL, total_mU_in_homogenate,
  activity_mU_per_mg_protein, nmol_in_window, within_std_range, cv_all,
  n_elevated, n, cv_flag, label_n [16x24]
mapping:  x = ~reorder(sample_id, activity_mU_per_uL), y = ~activity_mU_per_uL, fill = ~cv_flag
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: label = ~label_n 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge 

mapping: y = ~y 
geom_blank: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

Raw activity (mU/µL) spans a real 2.5-fold range across the sixteen individuals – 0.0547 (F07_01_ambient) to 0.1356 (F07_04_36C) – against a usable-replicate technical CV under 6.5% for every sample, so this spread reflects genuine inter-individual biological/extraction variation rather than measurement noise. Because no background correction could be applied (note 4), these values are the raw, uncorrected rates.

9.4 Protein-normalized activity

# 15 of 16 samples are "clean" here; F07_03_ambient (CV > 15% on all raw
# reps, note above) is excluded from the trend fit.
norm_clean <- activity_plot_data %>% filter(cv_flag == qual_levels[1])

norm_plot <- ggplot(activity_plot_data,
                    aes(x = total_protein_mg, y = activity_mU_per_mg_protein)) +
  geom_smooth(data = norm_clean, method = "lm", se = TRUE,
              colour = "grey35", fill = "grey85", linewidth = 0.6) +
  geom_point(aes(colour = cv_flag), size = 3.5) +
  geom_text(aes(label = individual), vjust = -1, size = 3, colour = "grey25") +
  scale_colour_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  labs(title = "Protein-normalized CS activity vs. total extracted protein",
       subtitle = paste0("Point labels = individual ID. Total protein = measured protein ",
                         "concentration x the same\n", assay_params$homogenate_volume_uL,
                         " uL homogenization volume used for every sample. Grey line = fit to ",
                         "15 of 16\nsamples (F07_03_ambient excluded, CV > 15% on raw reps); a ",
                         "positive raw-rate/protein\nrelationship would suggest normalization is ",
                         "correcting for real extraction-efficiency\ndifferences rather than ",
                         "introducing an artifact."),
       x = "Total extracted protein (mg)", y = "CS activity (mU/mg protein)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

protein_trend <- cor.test(activity_plot_data$total_protein_mg, activity_plot_data$activity_mU_per_mg_protein)
protein_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_mg_protein)
uL_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_uL)

ggsave(file.path(output_dir, "citrate_synthase_activity_normalized.png"), norm_plot,
       width = 10, height = 6, dpi = 300)

cat("Correlation of mU/mg protein with total extracted protein\n")
cat(sprintf("  all %d samples      : r = %+.3f, p = %.4f\n",
            nrow(activity_plot_data), protein_trend$estimate, protein_trend$p.value))
cat(sprintf("  clean samples only  : r = %+.3f, p = %.4f  (n = %d)\n",
            protein_trend_clean$estimate, protein_trend_clean$p.value, nrow(norm_clean)))
cat(sprintf("  clean-sample mU/mg protein range: %.3f to %.3f (%.2f-fold)\n",
            min(norm_clean$activity_mU_per_mg_protein), max(norm_clean$activity_mU_per_mg_protein),
            max(norm_clean$activity_mU_per_mg_protein) / min(norm_clean$activity_mU_per_mg_protein)))
cat(sprintf("  clean-sample mU/uL range: %.4f to %.4f (%.2f-fold)\n",
            min(norm_clean$activity_mU_per_uL), max(norm_clean$activity_mU_per_uL),
            max(norm_clean$activity_mU_per_uL) / min(norm_clean$activity_mU_per_uL)))
cat(sprintf("\nArtifact check: r(total protein, mU/uL raw rate) = %+.3f, p = %.4f  (n = %d)\n",
            uL_trend_clean$estimate, uL_trend_clean$p.value, nrow(norm_clean)))
cat("If mU/uL itself trends with total protein while mU/mg protein trends oppositely,\n",
    "the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate\n",
    "-- but a STRONG raw-rate/protein relationship (as seen here) instead argues that\n",
    "protein normalization is correcting for genuine extraction-efficiency differences.\n", sep = "")

cat("\n--- protein_trend_clean: correlation test on clean samples ---\n\n")
str(protein_trend_clean, max.level = 1, give.attr = FALSE)

cat("\n--- norm_clean: clean samples used for the trend fit ---\n\n")
str(norm_clean)

norm_plot

Correlation of mU/mg protein with total extracted protein
  all 16 samples      : r = -0.874, p = 0.0000
  clean samples only  : r = -0.868, p = 0.0000  (n = 15)
  clean-sample mU/mg protein range: 79.566 to 207.819 (2.61-fold)
  clean-sample mU/uL range: 0.0547 to 0.1356 (2.48-fold)

Artifact check: r(total protein, mU/uL raw rate) = +0.935, p = 0.0000  (n = 15)
If mU/uL itself trends with total protein while mU/mg protein trends oppositely,
the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate
-- but a STRONG raw-rate/protein relationship (as seen here) instead argues that
protein normalization is correcting for genuine extraction-efficiency differences.

--- protein_trend_clean: correlation test on clean samples ---

List of 9
 $ statistic  : Named num -6.32
 $ parameter  : Named int 13
 $ p.value    : num 2.67e-05
 $ estimate   : Named num -0.868
 $ null.value : Named num 0
 $ alternative: chr "two.sided"
 $ method     : chr "Pearson's product-moment correlation"
 $ data.name  : chr "norm_clean$total_protein_mg and norm_clean$activity_mU_per_mg_protein"
 $ conf.int   : num [1:2] -0.956 -0.642

--- norm_clean: clean samples used for the trend fit ---

tibble [15 × 24] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:15] "F07_01_36C" "F07_01_ambient" "F07_02_36C" "F07_02_ambient" ...
 $ n_reps_used               : int [1:15] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_rate_mOD_min         : num [1:15] 4.93 3.43 7.48 7.43 8.28 ...
 $ sd_rate                   : num [1:15] 0.0577 0.0577 0.1258 0.1443 0.2566 ...
 $ cv_rate                   : num [1:15] 1.17 1.68 1.68 1.94 3.1 ...
 $ family                    : chr [1:15] "F07" "F07" "F07" "F07" ...
 $ individual                : chr [1:15] "01" "01" "02" "02" ...
 $ temperature               : chr [1:15] "36C" "ambient" "36C" "ambient" ...
 $ conc_ug_mL                : num [1:15] 663 263 932 1080 1206 ...
 $ total_protein_mg          : num [1:15] 0.2322 0.0921 0.3263 0.378 0.4219 ...
 $ bg_rate_mOD_min           : num [1:15] 0 0 0 0 0 0 0 0 0 0 ...
 $ corrected_mOD_min         : num [1:15] 4.93 3.43 7.48 7.43 8.28 ...
 $ rate_OD_min               : num [1:15] 0.00493 0.00343 0.00748 0.00743 0.00828 ...
 $ nmol_per_min              : num [1:15] 0.157 0.109 0.238 0.237 0.264 ...
 $ activity_mU_per_uL        : num [1:15] 0.0786 0.0547 0.1192 0.1184 0.1319 ...
 $ total_mU_in_homogenate    : num [1:15] 27.5 19.1 41.7 41.4 46.2 ...
 $ activity_mU_per_mg_protein: num [1:15] 118 208 128 110 109 ...
 $ nmol_in_window            : num [1:15] 3.14 2.19 4.77 4.74 5.28 ...
 $ within_std_range          : logi [1:15] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ cv_all                    : num [1:15] 1.17 1.68 1.68 1.94 3.1 ...
 $ n_elevated                : int [1:15] 0 0 0 0 0 0 0 0 0 0 ...
 $ n                         : int [1:15] 3 3 3 3 3 3 3 3 3 3 ...
 $ cv_flag                   : Factor w/ 3 levels "clean","CV > 15%",..: 1 1 1 1 1 1 1 1 1 1 ...
 $ label_n                   : chr [1:15] "3/3 reps" "3/3 reps" "3/3 reps" "3/3 reps" ...

Raw activity (mU/µL) rises strongly and significantly with total extracted protein (r = +0.935, p = 0.0000), while mU/mg protein correlates negatively with total extracted protein (r = -0.868, p = 0.0000, n = 15, excluding F07_03_ambient). Because raw signal genuinely increases with more protein loaded – consistent with more CS enzyme being present when more protein was extracted – the negative mU/mg protein trend here is more consistent with protein normalization doing its intended job (correcting for real inter-sample extraction-efficiency differences) than with a pure division artifact. Normalizing here does not narrow the inter-individual spread – it is essentially unchanged, from a 2.48-fold raw range to a 2.61-fold range in mU/mg protein – so on this plate the case for normalization rests on the raw-rate/protein correlation itself, not on a visibly tighter normalized range.

10 QC SUMMARY

qc_summary <- build_qc_summary_table(layout_wells, recon$summary, std_curve,
                                      well_diagnostics, well_rates, cv_summary,
                                      baseline_per_sample, pos_control, assay_params)

kable(qc_summary, col.names = c("QC check", "Result"),
      caption = "Quality control summary for this plate")

write.csv(qc_summary, file.path(output_dir, "qc_summary.csv"), row.names = FALSE)

cat("\n--- qc_summary: consolidated QC checks for the run ---\n\n")
str(qc_summary)
QC check Result
Occupied wells in layout 72
Layout wells missing from absorbance CSV 0
Max disagreement between absorbance CSV and full report 0
Standard concentrations with replicate CV > threshold
Standard wells flagged as outliers 0/18 ()
Standard curve R^2 (outlier-excluded, replicate level) 0.9992
Standard curve R^2 (all wells, replicate level) 0.9992
Standard wells that LOST signal over the run 15/18
Standards above photometric linearity ceiling
Background control wells behaving as active reactions 3/3
Sample wells with a decreasing trace 1/48
Sample wells with an elevated starting A412 0/48 ()
Samples with ALL THREE replicates baseline-compromised 0/16 ()
Sample wells with a read glitch 0/48
Sample wells usable for rate extraction 47/48
Samples with technical CV > threshold (all reps) 1/16
Samples with technical CV > threshold (usable reps) 0/16
Positive control replicates rising and linear 3/3

Quality control summary for this plate

--- qc_summary: consolidated QC checks for the run ---

'data.frame':   18 obs. of  2 variables:
 $ check: chr  "Occupied wells in layout" "Layout wells missing from absorbance CSV" "Max disagreement between absorbance CSV and full report" "Standard concentrations with replicate CV > threshold" ...
 $ value: chr  "72" "0" "0" "" ...

10.1 Auto-generated findings

The bullets below are generated programmatically from the QC/CV/anomaly objects above by generate_qc_findings(), so the same logic that flags a problem here is guaranteed to be the logic that flagged it upstream.

qc_findings <- generate_qc_findings(well_diagnostics, well_rates, cv_summary,
                                     baseline_per_sample, std_curve, recon$summary,
                                     assay_params)

cat(paste0("- ", qc_findings, "\n"), sep = "\n")
- **Kinetic sources agree.** absorbance CSV and full report agree exactly on all 792 shared well x timepoint readings; all layout wells are present.

- **Decreasing kinetics.** 1 sample well(s) end lower than (or equal to) their starting A412 and are excluded from rate extraction: A7.

- **Background control wells behaving as active reactions.** 3 of 3 background wells show a starting A412 or drift inconsistent with a flat, inactive well: F10, F11, F12.

- **Standard signal decay.** 15 of 18 standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.

- **Standard curve fit (outlier-excluded):** slope = 0.03140 A412/nmol, R^2 = 0.9992 (all-wells R^2 = 0.9992).

- **Technical replicate CV > 15% (all replicates).** 1 of 16 samples: F07_03_ambient (87%).

11 SUMMARY

11.1 Anomalies found

  1. All three pooled-background replicates are compromised by a confirmed Reaction Mix / Background Control Mix mix-up (note 4). F10-F12 rise 0.067-0.119 A412 over the run – comparable to a real reaction, not a background – so compute_background_correction() cannot estimate a background rate. Every sample’s corrected rate on this plate equals its raw rate.
  2. One sample well is dead. A7 (F07_03_ambient) reads a constant A412 0.099 across all 11 timepoints (net change = 0, R2 = 0) rather than a slow reaction. Its two siblings (A8, A9) are normal, so F07_03_ambient’s activity is computed from 2 of 3 replicates with a 6.4% usable-replicate CV.
  3. Widespread but harmless GSH standard signal decay. All 15 non-zero-concentration standard wells lose signal between t = 0 and t = 20 min, consistent with TNB2- instability. Because the standard curve is calibrated at t = 0, this has no effect on the fit (R2 = 0.9992, zero flagged outlier wells).
  4. The positive control was clean on this plate (2.6% CV, R2 0.998-1.000), confirming the core reaction chemistry and reader were functioning normally.

11.2 Technical replicate precision

Every sample is within the 15% CV threshold on usable replicates. F07_03_ambient is the only sample whose all-replicates CV fails (86.9%, driven entirely by the dead A7 well); its usable-replicate CV is a normal 6.4%. Every other sample is under 3.1% CV on all three raw replicates, most under 2%.

11.3 Assay validity

The positive control and standard curve both worked well: all three positive-control replicates rise linearly (R2 > 0.997) with a tight 2.6% CV, and the GSH standard curve has zero flagged outlier wells (R2 = 0.9992). The background control did not work (note 4) – this plate’s activity values are raw, uncorrected rates, not background-corrected ones. Where a pooled background control has worked correctly, it has been a small term (roughly 2-7% of the sample signal), so the raw values here are likely only a modest overestimate of true background-corrected activity, but this cannot be confirmed without a working background control.

11.4 Recommendation

  1. Samples, standard curve, and positive control do not need re-assay. Only A7 (F07_03_ambient) failed outright, and that sample still has two clean, precise replicates.
  2. The background control should be re-run before treating this plate’s activity values as final. Consider physically separating the Reaction Mix and Background Control Mix reservoirs/tips, or pipetting the background wells first, before switching reagents for the sample wells, to reduce the risk of this mix-up recurring.
  3. Protein normalization looks appropriate for this plate, not artifactual: raw rate (mU/µL) is strongly and significantly correlated with total extracted protein (r = +0.93, p = 0.0000), so the negative mU/mg-protein trend is better explained by genuine extraction-efficiency differences between samples than by a division artifact.

11.5 What can be used from this plate

15 of 16 samples are fully clean, and the 16th (F07_03_ambient) is still usable via 2 of 3 replicates. Because this plate’s rates are uncorrected, they should not be pooled directly with background-corrected values from other assays without accounting for that difference. Within this plate, raw activity spans 0.0547-0.1356 mU/uL, a real 2.5-fold range against <6.5% technical CV per sample.

  • Re-running just the background triplicate (rather than the whole plate) would let this plate’s existing sample, standard-curve, and positive-control data be background-corrected without re-collecting everything else.
  • Protein-normalized activity (mU/mg protein) is the recommended basis for downstream comparison on this plate: raw activity correlates strongly with total extracted protein, consistent with normalization correcting a real extraction-efficiency covariate rather than introducing one.
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase ABCAM https://robertslab.github.io/sams-notebook/posts/2026/2026-08-24-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-7-Ambient-and-36C-Ctenidia/ Mon, 24 Aug 2026 07:00:00 GMT Citrate Synthase Assay - SORMI June 2026 M.gigas Family 5 36C Ctenidia Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-14-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-5-36C-Ctenidia/

INTRO

Continuing citrate synthase assays on Family 5 from the June 2026 SORMI experiment, I assayed ctenidia from the eight Magallana gigas (Pacific oyster) Family 5 individuals held at 36°C (heat-stress exposure) using the Citrate Synthase Assay Kit (ABCAM; cat: ab239712). Samples had been previously homogenized on 20260811 (Notebook entry), and protein concentrations for normalization were determined on 20260813 (Notebook entry). This is the companion plate to the Family 5 ambient assay run on 20260811 (Notebook entry).

Important

The plate reader assay time differed between this run and the 20260811 run.

Run Read duration Reads Interval
20260811 - Family 5 ambient 40 min 21 2 min
20260814 - Family 5 36°C (this run) 20 min 11 2 min

This plate was read for 20 minutes (11 reads), half the 40-minute (21-read) kinetic window used on 20260811. Both runs used 2-minute read intervals at 412 nm and 25°C, so the difference is purely in how long the kinetics were followed.

Rate extraction uses a fixed-width sliding window (5 points = 8 min) rather than the full trace, so the shorter read does not change the width of the window a rate is fit over — only how many candidate windows are available to slide across. It does change where the steepest window lands. On this plate every sample well was fitted on the window running from 12 to 20 min — the final five reads of the run. On 20260811, where traces rose immediately and then decelerated, the great majority of wells were fitted on the initial 0-8 min window.

The fits themselves are excellent (every sample well R2 > 0.99), so the reported rates are well determined over the window used. But since the steepest window here is the last one in the run, there is no read time after it to confirm the reaction had reached its maximum rate. Any ambient-vs-36°C comparison should note that the two plates did not sample the same portion of the kinetic curve, and read duration should be standardized before further families are run.

All data/code/analysis is available here:

MATERIALS & METHODS

Citrate Synthase Assay

The Citrate Synthase Assay Kit (ABCAM; cat: ab239712) was used according to the manufacturer’s instructions (ABCAM-Citrate-Synthase-Assay-v4a-ab239712.pdf (GitHub; PDF)).

Homogenates were thawed and 2 µL of undiluted homogenate was combined with 48 µL of Assay Buffer 7 per well, followed by 50 µL of Reaction Mix. Each sample was assayed in technical triplicate in a clear, flat-bottomed 96-well plate. Every sample also received a paired triplicate of background control wells, which receive Background Control Mix (no CS Substrate Mix) to correct for pre-existing free thiol in the homogenate. A six-point GSH standard curve (0, 8, 16, 24, 32, 40 nmol/well) and a CS positive control were included in triplicate on the same plate.

Master Mix Calcs

REACTION MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 43 50 2150
DTNB 5 50 250
CS Substrate 2 50 100

BACKGROUND MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 45 30 1350
DTNB 5 30 150

Absorbance was read at 412 nm in kinetic mode at 25°C on a Synergy HTX plate reader (Agilent), at 2-minute intervals. This plate was read for 20 minutes (11 reads); see the callout above.

Normalization

Activity was normalized to total extracted protein rather than tissue weight. Tissue was homogenized in 350 µL of Assay Buffer 7, and protein concentration in each homogenate was quantified separately on 20260813 (Notebook entry) using the Bradford assay. Total protein mass per sample is that concentration multiplied by the 350 µL homogenization volume. Tissue weights are recorded in the plate layout labels for provenance only and play no role in the activity calculation.

Analysis

Data were analyzed in R. The Rmd file is here (GitHub):

Knitted markdown is below.

Data files

Plate reader files

Plate reader files (*.xpt) are available here (require Gen5 software to open):

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/plate_reader_files

Raw absorbance data

Raw absorbance data (*.csv), plate layouts, and Gen5 full reports are available here:

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/raw_absorbance

RESULTS

This plate ran cleanly. All 24 sample wells rise monotonically with replicate-level R2 > 0.99, no well required exclusion, all eight samples pass the 15% technical CV threshold (highest is F05_08_36C at 7.8%), and the positive control was linear at 2.8-3.0 mA412/min. The GSH standard curve fails QC only by a single clearly-identified outlier well (F5, 40 nmol); excluding it drops that concentration’s CV from 15.9% to 0.5% and lifts the replicate-level R2 to 0.9985. This plate does not need re-assay.

Background-corrected rates cluster tightly — 2.13 to 2.30 mA412/min across all eight individuals, a spread of only 7.8% against 1-8% technical CV per sample. No striking within-family variation in raw CS activity is apparent at 36°C.

Sample Reps used CV all reps (%) Corrected rate (mA412/min) Activity (mU/mg protein) Interpretation
F05_01_36C 3/3 3.3 2.17 37.0 usable
F05_02_36C 3/3 2.4 2.22 66.0 usable
F05_03_36C 3/3 3.5 2.30 56.0 usable
F05_04_36C 3/3 2.3 2.28 35.0 usable
F05_05_36C 3/3 1.2 2.15 21.0 usable
F05_06_36C 3/3 2.4 2.13 27.4 usable
F05_07_36C 3/3 1.2 2.20 32.2 usable
F05_08_36C 3/3 7.8 2.25 26.9 usable

Two items still need follow-up. One background well (C8) showed the signature of a Reaction Mix / Background Control Mix swap — the same dispensing issue seen more extensively on the 20260811 plate — and is dropped automatically, leaving F05_03_36C with 2 of 3 background replicates. Separately, raw rate (mU/µL) is essentially flat against total extracted protein here (r = -0.48, p = 0.23) while mU/mg protein trends negatively with it, which points to the normalized figures reflecting a division artifact rather than a biological effect. Raw rate is the safer basis for downstream comparison until a dilution series resolves this.

AI Use Level 4: Substantial AI contribution

1 BACKGROUND

Citrate synthase (CS) activity in ctenidia of eight Magallana gigas (Pacific oyster) individuals from family F05 held at 36 °C (heat-stress exposure), assayed 2026-08-14 with the Abcam Citrate Synthase Assay Kit (ab239712), v4a. The kit is a coupled kinetic assay. CS condenses acetyl-CoA and oxaloacetate, releasing free CoA-SH; the liberated thiol reduces DTNB to TNB2-, which absorbs at 412 nm. The rate of A412 increase is proportional to CS activity, and absolute nmol of thiol are assigned from a GSH (reduced glutathione) standard curve read on the same plate.

Absorbance was read at 412 nm in kinetic mode, 25 °C, every 2 min for 20 min (11 reads) on a Synergy HTX. See note 1.

This document is fully self-contained: layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are all defined directly below (see ## Pipeline functions under SETUP) rather than sourced from a separate script, so the whole analysis can be reproduced from this one file.

1.1 Sample naming convention

Well labels follow <family>_<individual>_<temperature>[_BG]-<assay_type>-<weight>-df.<n> for sample/background wells (e.g. F05_01_36C-citrate_synthase-12.8-df.0), STD-<assay_type>-<nmol_per_well> for GSH standards, and POS-<assay_type> for the positive control, with the temperature exposure encoded in the well label as _36C. The trailing number in the sample label (e.g. 12.8) is tissue weight in mg, which is not used for normalization here (note 2); it is parsed and validated but plays no role in the activity calculation.

1.2 Notes

  1. Read duration for this plate is 20 minutes (11 reads at 2-min intervals). The assay parameters below set read_duration_min = 20 to match. Rate extraction uses a fixed-width sliding window (rate_window_n = 5 points = 8 min) rather than the full trace, so this only changes how many windows are available to slide across; see the KINETIC TRACES section for whether the assay is still in its linear phase over this window.
  2. Activity is normalized to total extracted protein, not tissue weight, because protein concentration was measured directly (BCA-style assay, sample_protein_concentrations_*.csv), and controls for extraction efficiency in a way tissue weight cannot.
  3. Protein concentrations are read from both supplied CSV files. This plate’s F05_07_36C record is only present in file 1, and under a slightly different string (F-05_07_36C, extra hyphen) — corrected via sample_id_fixes in the protein-concentration chunk.
library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

1.3 Pipeline functions

All layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are defined directly in this document (rather than sourced from a separate script) so that it is fully self-contained and reproducible from this single file. The functions are grouped below by pipeline stage; each is called in the DATA / KINETIC TRACES / ANOMALY DETECTION / GSH STANDARD CURVE / RATE EXTRACTION / BACKGROUND CORRECTION / TECHNICAL REPLICATE PRECISION / CITRATE SYNTHASE ACTIVITY / QC SUMMARY sections that follow. Nothing plate-specific (typo fixes, disqualified samples, thresholds) is hardcoded in any function – it is always passed in as an argument, from the assay_params list defined further below or from a plate-specific fix list.

1.3.1 Data import and layout parsing

Reads the raw Gen5 exports (kinetic absorbance CSV, full-report text backup, plate-layout CSV), reconciles them against each other, and attaches well-type/sample-id/replicate metadata plus measured protein concentrations to every well.

## -----------------------------------------------------------------------
## elapsed_to_min()
## Purpose : Convert Gen5 "H:MM:SS" elapsed-time strings to decimal minutes.
## Inputs  : x - character vector of "H:MM:SS" strings
## Outputs : numeric vector of elapsed minutes, same length as x
## -----------------------------------------------------------------------
elapsed_to_min <- function(x) {
  vapply(strsplit(x, ":"), function(p) {
    p <- as.numeric(p)
    p[1] * 60 + p[2] + p[3] / 60
  }, numeric(1))
}

## -----------------------------------------------------------------------
## read_absorbance_csv()
## Purpose : Parse a Gen5 "absorbance-*.csv" kinetic export: wavelength
##           header, a "Time," row, a temperature row, then one row per well
##           (well ID in column 1, one OD reading per timepoint thereafter).
## Inputs  : path - path to the absorbance-*.csv file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_absorbance_csv <- function(path) {
  ln <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  time_i   <- grep("^Time,", ln)[1]
  time_str <- strsplit(ln[time_i], ",")[[1]]
  time_str <- time_str[!time_str %in% c("", "Time")]
  tmin     <- elapsed_to_min(time_str)

  well_ln <- ln[grepl("^[A-H][0-9]{1,2},", ln)]
  bind_rows(lapply(strsplit(well_ln, ","), function(p) {
    data.frame(well = p[1], time_min = tmin,
               od = as.numeric(p[seq_along(tmin) + 1]),
               stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## read_full_report()
## Purpose : Parse a Gen5 "full_report-*.txt" export: tab-delimited, wells
##           across columns, timepoints down rows. Used only as an
##           independent cross-check against the absorbance CSV, never as a
##           primary data source.
## Inputs  : path - path to the full_report-*.txt file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_full_report <- function(path) {
  ln  <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  hdr_i <- grep("^Time\t.*\tA1\t", ln)[1]
  hdr   <- trimws(strsplit(ln[hdr_i], "\t")[[1]])
  rows  <- ln[grepl("^[0-9]:[0-9]{2}:[0-9]{2}\t", ln)]
  m     <- do.call(rbind, strsplit(rows, "\t"))
  tmin  <- elapsed_to_min(m[, 1])

  well_cols <- which(grepl("^[A-H][0-9]{1,2}$", hdr))
  bind_rows(lapply(well_cols, function(j) {
    data.frame(well = hdr[j], time_min = tmin,
               od = as.numeric(m[, j]), stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## parse_plate_layout()
## Purpose : Convert a raw Gen5 plate-layout CSV export into one row per
##           occupied well with its descriptive label. Auto-detects two
##           export formats seen in practice:
##             "single"  - one row per plate row (row 1 = column numbers,
##                         column 1 = row letter, last column = "Name" tag);
##                         each cell already holds the full descriptive label.
##             "double"  - each plate row is exported as TWO rows: a
##                         "Well ID" row (SPL codes) immediately followed by
##                         a "Name" row (the descriptive label actually used
##                         downstream).
##           The format is detected from column 1: if a value repeats twice
##           in a row (once for the well-ID row, once for the label row)
##           immediately below a numeric header, it is treated as "double".
## Inputs  : path - path to the layout-*.csv file
## Outputs : data.frame, one row per occupied well:
##           well (chr, e.g. "A1"), plate_row (chr), plate_col (int),
##           label (chr, trimmed descriptive label)
## -----------------------------------------------------------------------
parse_plate_layout <- function(path) {
  plate_layout <- read.csv(path, header = FALSE, stringsAsFactors = FALSE)

  col1 <- trimws(plate_layout$V1)
  # "double" format: row letters appear twice in a row in column 1
  # (once tagging the Well-ID row, once tagging the Name row), e.g.
  # "A", "A", "B", "B", ... Detect by checking whether non-empty,
  # non-header values in col1 are each immediately followed by a repeat.
  data_rows   <- which(col1 != "" & !is.na(col1))[-1]  # drop header row (row 1)
  is_double <- length(data_rows) >= 2 &&
    all(col1[data_rows[c(TRUE, FALSE)]] == col1[data_rows[c(FALSE, TRUE)]][seq_len(floor(length(data_rows) / 2))])

  if (is_double) {
    # Pair each row-letter row with the following label row.
    row_letter_rows <- data_rows[c(TRUE, FALSE)]
    label_rows      <- data_rows[c(FALSE, TRUE)]
    stopifnot(length(row_letter_rows) == length(label_rows))

    out <- expand.grid(pair_idx = seq_along(row_letter_rows),
                        col_idx  = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = col1[row_letter_rows[pair_idx]],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j],
                                   label_rows[pair_idx], col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  } else {
    # "single" format: row 1 = column numbers, column 1 = row letters,
    # last column = "Name" tag; each cell already holds the full label.
    out <- expand.grid(row_idx = 2:nrow(plate_layout),
                        col_idx = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = plate_layout$V1[row_idx],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j], row_idx, col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  }

  attr(out, "layout_format") <- ifelse(is_double, "double", "single")
  out
}

## -----------------------------------------------------------------------
## reconcile_kinetic_sources()
## Purpose : Cross-check the absorbance CSV (primary data source) against
##           the full_report text export (independent instrument export)
##           and the plate layout (expected well set). Fails loudly
##           (stopifnot) if any layout well is missing from the CSV, or if
##           the two kinetic sources disagree on any shared reading.
## Inputs  : absorbance_csv - long-format data.frame from read_absorbance_csv()
##           full_report    - long-format data.frame from read_full_report()
##           layout_wells   - data.frame from parse_plate_layout()
##           tolerance      - max allowed |CSV - report| absolute difference
##                            (default 1e-9, i.e. exact agreement)
## Outputs : list(plate_readings, overlap_check, summary), where
##           plate_readings = absorbance_csv with a `source` column added,
##           summary        = named list of counts for reporting
## -----------------------------------------------------------------------
reconcile_kinetic_sources <- function(absorbance_csv, full_report, layout_wells,
                                       tolerance = 1e-9) {
  wells_csv    <- unique(absorbance_csv$well)
  wells_report <- unique(full_report$well)

  wells_report_not_csv <- setdiff(wells_report, wells_csv)
  wells_layout_not_csv <- setdiff(layout_wells$well, wells_csv)
  wells_csv_not_layout <- setdiff(wells_csv, layout_wells$well)

  overlap_check <- inner_join(absorbance_csv, full_report,
                              by = c("well", "time_min"), suffix = c("_csv", "_report")) %>%
    mutate(abs_diff = abs(od_csv - od_report))

  max_disagreement <- if (nrow(overlap_check) > 0) max(overlap_check$abs_diff) else NA_real_

  stopifnot(
    length(wells_layout_not_csv) == 0,
    is.na(max_disagreement) || max_disagreement < tolerance
  )

  plate_readings <- absorbance_csv %>% mutate(source = "absorbance_csv")

  list(
    plate_readings = plate_readings,
    overlap_check  = overlap_check,
    summary = list(
      n_wells_csv               = length(wells_csv),
      n_wells_report            = length(wells_report),
      n_wells_report_not_csv    = length(wells_report_not_csv),
      n_wells_layout_not_csv    = length(wells_layout_not_csv),
      n_wells_csv_not_layout    = length(wells_csv_not_layout),
      n_shared_readings         = nrow(overlap_check),
      max_disagreement          = max_disagreement
    )
  )
}

## -----------------------------------------------------------------------
## annotate_wells()
## Purpose : Join layout labels onto the kinetic readings and parse each
##           label into well_type, sample_id, family, individual,
##           temperature, tissue weight, and standard concentration.
##           Expected label grammar:
##             sample/background : <family>_<individual>_<temperature>[_BG]-<assay_type>-<weight>-df.<n>
##             standard           : STD-<assay_type>-<nmol_per_well>
##             positive control   : POS-<assay_type>
##           Label typos (e.g. a misspelled assay_type) are corrected before
##           parsing via `label_fixes`, a named character vector of
##           c(pattern = replacement) pairs applied in order with gsub().
## Inputs  : plate_readings - data.frame from reconcile_kinetic_sources()$plate_readings
##           layout_wells   - data.frame from parse_plate_layout()
##           label_fixes    - optional named character vector, e.g.
##                            c("citrate_synthasse" = "citrate_synthase")
## Outputs : plate_long - fully annotated long-format data.frame, one row
##           per well x timepoint, with well_type, sample_id, family,
##           individual, temperature, weight_mg, std_nmol columns added
## -----------------------------------------------------------------------
annotate_wells <- function(plate_readings, layout_wells, label_fixes = NULL) {
  plate_long <- plate_readings %>%
    left_join(layout_wells, by = "well") %>%
    mutate(
      well_type = case_when(
        grepl("^STD-", label) ~ "standard",
        grepl("^POS-", label) ~ "positive_control",
        grepl("_BG-",  label) ~ "background",
        TRUE                  ~ "sample"
      ),
      label_clean = label
    )

  if (!is.null(label_fixes)) {
    for (pat in names(label_fixes)) {
      plate_long$label_clean <- gsub(pat, label_fixes[[pat]], plate_long$label_clean)
    }
  }

  plate_long <- plate_long %>%
    mutate(
      sample_id   = ifelse(well_type %in% c("sample", "background"),
                           sub("(_BG)?-[a-zA-Z_]+-.*$", "", label_clean), NA_character_),
      family      = ifelse(!is.na(sample_id), sub("^(F[0-9]+)_.*$", "\\1", sample_id), NA_character_),
      individual  = ifelse(!is.na(sample_id), sub("^F[0-9]+_([0-9]+)_.*$", "\\1", sample_id), NA_character_),
      temperature = ifelse(!is.na(sample_id), sub("^F[0-9]+_[0-9]+_(.*)$", "\\1", sample_id), NA_character_),
      weight_mg   = ifelse(well_type %in% c("sample", "background"),
                           as.numeric(sub("^.*-[a-zA-Z_]+-([0-9.]+)-df\\..*$", "\\1", label_clean)),
                           NA_real_),
      std_nmol    = ifelse(well_type == "standard",
                           as.numeric(sub("^STD-[a-zA-Z_]+-", "", label_clean)), NA_real_)
    ) %>%
    arrange(well_type, sample_id, well, time_min)

  # Fail loudly rather than silently dropping malformed labels
  stopifnot(
    !any(is.na(plate_long$well_type)),
    !any(is.na(plate_long$weight_mg[plate_long$well_type %in% c("sample", "background")])),
    !any(is.na(plate_long$std_nmol[plate_long$well_type == "standard"]))
  )

  plate_long
}

## -----------------------------------------------------------------------
## load_protein_concentrations()
## Purpose : Load one or more BCA/Bradford-style protein-concentration CSV
##           exports, concatenate them, and match each plate sample to
##           exactly one protein-concentration record. Fails loudly if any
##           plate sample has zero or more than one match. Converts
##           concentration (ug/mL) to total protein mass (mg) in the full
##           homogenate using the measured homogenization volume.
## Inputs  : protein_files        - character vector of CSV paths, each with
##                                   columns `Sample` and
##                                   `Calculated concentration (ug/mL)`
##           plate_sample_ids     - character vector of this plate's sample IDs
##           homogenate_volume_uL - measured homogenization volume (uL),
##                                   applied identically to every sample
##           sample_id_fixes      - optional named character vector of
##                                   c(pattern = replacement) label-typo
##                                   corrections applied to the protein
##                                   file's Sample column before matching,
##                                   e.g. c("F-05_07_36C" = "F05_07_36C")
## Outputs : data.frame, one row per plate sample:
##           sample_id, source_file, conc_ug_mL, total_protein_mg
## -----------------------------------------------------------------------
load_protein_concentrations <- function(protein_files, plate_sample_ids,
                                         homogenate_volume_uL,
                                         sample_id_fixes = NULL) {
  protein_all <- protein_files %>%
    purrr::map_dfr(~ read.csv(.x, check.names = FALSE) %>%
                      mutate(source_file = basename(.x))) %>%
    rename(sample_id_protein = Sample, conc_ug_mL = `Calculated concentration (ug/mL)`) %>%
    select(sample_id_protein, source_file, conc_ug_mL)

  if (!is.null(sample_id_fixes)) {
    for (pat in names(sample_id_fixes)) {
      protein_all$sample_id_protein <- gsub(pat, sample_id_fixes[[pat]],
                                             protein_all$sample_id_protein)
    }
  }

  protein_matches <- protein_all %>% filter(sample_id_protein %in% plate_sample_ids)
  match_counts    <- protein_matches %>% count(sample_id_protein, name = "n_matches")

  stopifnot(
    all(plate_sample_ids %in% protein_matches$sample_id_protein),
    all(match_counts$n_matches == 1)
  )

  protein_matches %>%
    rename(sample_id = sample_id_protein) %>%
    mutate(total_protein_mg = conc_ug_mL * homogenate_volume_uL / 1e6) %>%
    arrange(sample_id)
}

1.3.2 Well-level kinetic QC

Flags read glitches, non-rising (‘decreasing’) traces, elevated t0 baselines, and other well-level anomalies before any rate is trusted.

## -----------------------------------------------------------------------
## local_step_excess()
## Purpose : For a kinetic trace, measure how much each read-to-read step
##           departs from the AVERAGE of its two immediate neighbouring
##           steps. A large local excess flags a read glitch; comparing to
##           local neighbours (rather than the trace's overall median step)
##           tolerates traces that smoothly decelerate throughout.
## Inputs  : od - numeric vector of OD readings in time order
## Outputs : numeric vector of length length(od) - 1 (one per step)
## -----------------------------------------------------------------------
local_step_excess <- function(od) {
  d <- diff(od)
  n <- length(d)
  local_expect <- vapply(seq_len(n), function(i) {
    mean(d[intersect(c(i - 1L, i + 1L), seq_len(n))])
  }, numeric(1))
  d - local_expect
}

## -----------------------------------------------------------------------
## compute_well_diagnostics()
## Purpose : Compute four independent per-well trace-shape diagnostics from
##           the raw kinetic trace, before any rate fitting: direction
##           (net change), monotonicity (fraction of rising steps),
##           over-range (any read above the photometric linearity ceiling),
##           and discontinuity (a read-to-read glitch via local_step_excess).
##           Also flags an elevated starting absorbance for sample wells,
##           and background wells that behave like active reactions.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with od_linear_max, glitch_excess_od,
##                          sample_baseline_od, bg_flat_od_max,
##                          bg_flat_drift_max
## Outputs : well_diagnostics - one row per well, with od_first, od_last,
##           od_max, net_change, frac_rising, max_step, typical_step,
##           step_excess, glitch_at_min, and flag_* / n_flags columns
## -----------------------------------------------------------------------
compute_well_diagnostics <- function(plate_long, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, plate_row, plate_col, well_type, sample_id, label, std_nmol, source) %>%
    summarise(
      od_first        = first(od),
      od_last         = last(od),
      od_max          = max(od),
      net_change      = last(od) - first(od),
      frac_rising     = mean(diff(od) > 0),
      max_step        = diff(od)[which.max(abs(diff(od)))],
      typical_step    = median(abs(diff(od))),
      step_excess     = local_step_excess(od)[which.max(abs(local_step_excess(od)))],
      glitch_at_min   = time_min[which.max(abs(local_step_excess(od))) + 1L],
      .groups = "drop"
    ) %>%
    mutate(
      step_ratio      = abs(max_step) / pmax(typical_step, 0.001),
      flag_decreasing = net_change <= 0,
      flag_over_range = od_max > assay_params$od_linear_max,
      flag_glitch     = abs(step_excess) > assay_params$glitch_excess_od,
      flag_high_baseline = well_type == "sample" & od_first > assay_params$sample_baseline_od,
      flag_bg_active  = well_type == "background" &
                        (od_first > assay_params$bg_flat_od_max |
                         abs(net_change) > assay_params$bg_flat_drift_max),
      n_flags         = flag_decreasing + flag_over_range + flag_glitch +
                        flag_bg_active + flag_high_baseline
    )
}

## -----------------------------------------------------------------------
## compute_baseline_diagnostics()
## Purpose : Classify sample/background wells by starting absorbance
##           (elevated vs. normal) and summarize, per sample, how many of
##           its replicates are baseline-compromised. A sample with all
##           replicates elevated has no clean replicate at all.
## Inputs  : well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with sample_baseline_od
## Outputs : list(baseline_check, baseline_per_sample)
## -----------------------------------------------------------------------
compute_baseline_diagnostics <- function(well_diagnostics, assay_params) {
  baseline_check <- well_diagnostics %>%
    filter(well_type %in% c("sample", "background")) %>%
    mutate(baseline = ifelse(od_first > assay_params$sample_baseline_od,
                             "elevated", "normal")) %>%
    arrange(desc(od_first))

  baseline_per_sample <- baseline_check %>%
    filter(well_type == "sample") %>%
    group_by(sample_id) %>%
    summarise(n_elevated = sum(baseline == "elevated"), n = n(),
              median_baseline = median(od_first), .groups = "drop") %>%
    arrange(desc(n_elevated))

  list(baseline_check = baseline_check, baseline_per_sample = baseline_per_sample)
}

1.3.3 Standard curve and rate extraction

Fits the GSH standard curve, extracts each well’s rate via a sliding-window linear regression gated on R-squared, and summarizes the positive control.

## -----------------------------------------------------------------------
## fit_standard_curve()
## Purpose : Extract t=0 GSH standard readings, quantify signal drift over
##           the run, compute per-concentration replicate statistics, flag
##           outlier wells by deviation from their triplicate median, and
##           fit three candidate calibration lines (all wells, per-
##           concentration means, outlier-excluded). The outlier-excluded
##           fit is the calibration used downstream.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with std_outlier_od
## Outputs : list with standards_t0, standard_drift, standard_summary,
##           standards_flagged, fit_all_wells, fit_means, fit_no_outliers,
##           fit_comparison, std_slope, std_intercept, std_r2, std_nmol_max
## -----------------------------------------------------------------------
fit_standard_curve <- function(plate_long, assay_params) {
  standards_all <- plate_long %>%
    filter(well_type == "standard") %>%
    select(well, std_nmol, time_min, od, source)

  standards_t0 <- standards_all %>%
    filter(time_min == 0) %>%
    select(well, std_nmol, od, source) %>%
    arrange(std_nmol, well)

  standard_drift <- standards_all %>%
    group_by(well, std_nmol) %>%
    summarise(od_t0 = od[time_min == 0], od_t40 = od[time_min == max(time_min)],
              drift = od_t40 - od_t0, .groups = "drop") %>%
    arrange(std_nmol, well)

  standard_summary <- standards_t0 %>%
    group_by(std_nmol) %>%
    summarise(n = n(), mean_od = mean(od), sd_od = sd(od),
              se_od = sd(od) / sqrt(n()), cv_pct = 100 * sd(od) / mean(od),
              median_od = median(od), .groups = "drop") %>%
    arrange(std_nmol) %>%
    mutate(net_od = mean_od - mean_od[std_nmol == 0],
           od_per_nmol = ifelse(std_nmol > 0, net_od / std_nmol, NA_real_))

  standards_flagged <- standards_t0 %>%
    group_by(std_nmol) %>%
    mutate(triplicate_median = median(od),
           deviation = od - triplicate_median,
           is_outlier = abs(deviation) > assay_params$std_outlier_od) %>%
    ungroup() %>%
    arrange(desc(abs(deviation)))

  fit_all_wells   <- lm(od ~ std_nmol, data = standards_flagged)
  fit_means       <- lm(mean_od ~ std_nmol, data = standard_summary)
  fit_no_outliers <- lm(od ~ std_nmol, data = standards_flagged %>% filter(!is_outlier))

  fit_comparison <- bind_rows(
    data.frame(fit = "all wells",          n = nobs(fit_all_wells),
               slope = coef(fit_all_wells)[2],   intercept = coef(fit_all_wells)[1],
               r_squared = summary(fit_all_wells)$r.squared),
    data.frame(fit = "concentration means", n = nobs(fit_means),
               slope = coef(fit_means)[2],       intercept = coef(fit_means)[1],
               r_squared = summary(fit_means)$r.squared),
    data.frame(fit = "outlier-excluded",   n = nobs(fit_no_outliers),
               slope = coef(fit_no_outliers)[2], intercept = coef(fit_no_outliers)[1],
               r_squared = summary(fit_no_outliers)$r.squared)
  )

  list(
    standards_all      = standards_all,
    standards_t0       = standards_t0,
    standard_drift     = standard_drift,
    standard_summary   = standard_summary,
    standards_flagged  = standards_flagged,
    fit_all_wells      = fit_all_wells,
    fit_means          = fit_means,
    fit_no_outliers    = fit_no_outliers,
    fit_comparison     = fit_comparison,
    std_slope          = unname(coef(fit_no_outliers)[2]),
    std_intercept      = unname(coef(fit_no_outliers)[1]),
    std_r2             = summary(fit_no_outliers)$r.squared,
    std_nmol_max       = max(standard_summary$std_nmol)
  )
}

## -----------------------------------------------------------------------
## window_slopes()
## Purpose : Fit every sliding window of `w` consecutive timepoints on one
##           kinetic trace and return the slope (mOD/min) and R^2 of each.
## Inputs  : t - numeric vector of timepoints (min)
##           y - numeric vector of OD readings, same length as t
##           w - window width in number of points
## Outputs : data.frame, one row per window: t_start, t_end,
##           slope_mOD_min, r2
## -----------------------------------------------------------------------
window_slopes <- function(t, y, w) {
  n <- length(y)
  bind_rows(lapply(seq_len(n - w + 1), function(s) {
    i  <- s:(s + w - 1)
    f  <- lm(y[i] ~ t[i])
    r2 <- summary(f)$r.squared
    data.frame(t_start = t[i[1]], t_end = t[i[w]],
               slope_mOD_min = unname(coef(f)[2]) * 1000,
               r2 = ifelse(is.na(r2), 0, r2))
  }))
}

## -----------------------------------------------------------------------
## compute_well_rates()
## Purpose : For every well, scan all sliding windows of assay_params$
##           rate_window_n consecutive reads and record the max-increasing
##           window (used for activity) and the max-absolute window
##           (diagnostic only -- matches Gen5's own `Max V` convention,
##           which can report large negative "rates" for degrading wells).
##           A rate is `rate_usable` only if its R^2 clears the floor, the
##           well's net change is positive, and no read glitch falls inside
##           the fitted window.
## Inputs  : plate_long       - data.frame from annotate_wells()
##           well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with rate_window_n, rate_min_r2
## Outputs : well_rates - one row per well: t_start, t_end, slope_mOD_min,
##           r2, max_abs_slope_mOD_min, abs_window_is_negative,
##           glitch_in_window, rate_usable (plus diagnostic columns joined in)
## -----------------------------------------------------------------------
compute_well_rates <- function(plate_long, well_diagnostics, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, well_type, sample_id, std_nmol) %>%
    group_modify(~ {
      w   <- window_slopes(.x$time_min, .x$od, assay_params$rate_window_n)
      inc <- w[which.max(w$slope_mOD_min), ]
      abs_ <- w[which.max(abs(w$slope_mOD_min)), ]
      data.frame(
        t_start = inc$t_start, t_end = inc$t_end,
        slope_mOD_min = inc$slope_mOD_min, r2 = inc$r2,
        max_abs_slope_mOD_min = abs_$slope_mOD_min,
        abs_window_is_negative = abs_$slope_mOD_min < 0
      )
    }) %>%
    ungroup() %>%
    left_join(well_diagnostics %>% select(well, net_change, frac_rising, od_max,
                                          flag_decreasing, flag_over_range,
                                          flag_glitch, glitch_at_min,
                                          flag_bg_active, n_flags),
              by = "well") %>%
    mutate(
      glitch_in_window = flag_glitch & glitch_at_min > t_start & glitch_at_min <= t_end,
      rate_usable      = r2 >= assay_params$rate_min_r2 & net_change > 0 & !glitch_in_window
    )
}

## -----------------------------------------------------------------------
## compute_positive_control()
## Purpose : Compute the CS Positive Control's rate and activity, and check
##           that its replicates rose linearly -- the plate's proof that the
##           reaction chemistry (DTNB, coupling, reader) worked, independent
##           of any sample-specific problem.
## Inputs  : well_rates   - data.frame from compute_well_rates()
##           std_slope    - calibration slope (A412/nmol), from fit_standard_curve()
##           assay_params - list with sample_volume_uL, dilution_factor
## Outputs : pos_control - one row per positive-control well: t_start, t_end,
##           slope_mOD_min, r2, net_change, activity_mU_uL
## -----------------------------------------------------------------------
compute_positive_control <- function(well_rates, std_slope, assay_params) {
  well_rates %>%
    filter(well_type == "positive_control") %>%
    mutate(activity_mU_uL = (slope_mOD_min / 1000) / std_slope /
                            assay_params$sample_volume_uL * assay_params$dilution_factor)
}

1.3.4 Background correction, replicate CV, and activity calculation

Estimates background rate from flat background wells, computes technical-replicate coefficients of variation, and converts corrected rates into protein-normalized CS activity.

## -----------------------------------------------------------------------
## compute_background_correction()
## Purpose : Estimate each sample's background rate from its flat
##           (well-behaved) background-control replicates only, since
##           anomalous background wells do not measure background.
## Inputs  : well_rates - data.frame from compute_well_rates()
## Outputs : background_per_sample - one row per sample: n_bg_total,
##           n_bg_flat, bg_rate_flat (NA if no flat replicate exists),
##           bg_rate_all
## -----------------------------------------------------------------------
compute_background_correction <- function(well_rates) {
  well_rates %>%
    filter(well_type == "background") %>%
    group_by(sample_id) %>%
    summarise(
      n_bg_total    = n(),
      n_bg_flat     = sum(!flag_bg_active),
      bg_rate_flat  = ifelse(sum(!flag_bg_active) > 0,
                             mean(slope_mOD_min[!flag_bg_active]), NA_real_),
      bg_rate_all   = mean(slope_mOD_min),
      .groups = "drop"
    )
}

## -----------------------------------------------------------------------
## compute_replicate_cv()
## Purpose : Compute technical-replicate coefficient of variation (CV%) on
##           the extracted rate, both across ALL three replicates and
##           across USABLE replicates only (rate_usable == TRUE). Flags any
##           sample whose CV exceeds assay_params$cv_threshold_pct.
## Inputs  : well_rates       - data.frame from compute_well_rates()
##           sample_id_order  - character vector giving the row order /
##                              complete sample-ID set to report (typically
##                              the plate's sample IDs, e.g. from
##                              protein_by_sample$sample_id)
##           assay_params     - list with cv_threshold_pct
## Outputs : cv_summary - one row per sample: n_all, mean_all, sd_all,
##           cv_all, n_usable, mean_usable, sd_usable, cv_usable,
##           excluded_wells, fails_cv_all, fails_cv_usable
## -----------------------------------------------------------------------
compute_replicate_cv <- function(well_rates, sample_id_order, assay_params) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  cv_all <- sample_rates %>%
    group_by(sample_id) %>%
    summarise(n_all = n(), mean_all = mean(slope_mOD_min), sd_all = sd(slope_mOD_min),
              cv_all = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  cv_usable <- sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_usable = n(), mean_usable = mean(slope_mOD_min),
              sd_usable = sd(slope_mOD_min),
              cv_usable = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  data.frame(sample_id = sample_id_order, stringsAsFactors = FALSE) %>%
    left_join(cv_all, by = "sample_id") %>%
    left_join(cv_usable, by = "sample_id") %>%
    mutate(
      excluded_wells = vapply(sample_id, function(s) {
        w <- sample_rates$well[sample_rates$sample_id == s & !sample_rates$rate_usable]
        if (length(w) == 0) "-" else paste(w, collapse = ", ")
      }, character(1)),
      fails_cv_all    = cv_all > assay_params$cv_threshold_pct,
      fails_cv_usable = !is.na(cv_usable) & cv_usable > assay_params$cv_threshold_pct
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## calculate_cs_activity()
## Purpose : Compute per-sample CS activity following Abcam sec 10.3, using
##           only usable replicates, corrected for background rate, scaled
##           through the standard curve, and normalized to total extracted
##           protein (rather than tissue weight).
## Inputs  : well_rates            - data.frame from compute_well_rates()
##           protein_by_sample     - data.frame from load_protein_concentrations()
##           background_per_sample - data.frame from compute_background_correction()
##           std_slope             - calibration slope (A412/nmol), from fit_standard_curve()
##           std_nmol_max          - max calibrated GSH concentration, from fit_standard_curve()
##           assay_params          - list with sample_volume_uL, dilution_factor,
##                                    homogenate_volume_uL, read_duration_min
##           plate_long            - data.frame from annotate_wells(), used to
##                                    attach family/individual/temperature
## Outputs : cs_activity - one row per sample: n_reps_used, mean_rate_mOD_min,
##           sd_rate, cv_rate, family, individual, temperature, conc_ug_mL,
##           total_protein_mg, bg_rate_mOD_min, corrected_mOD_min,
##           activity_mU_per_uL, total_mU_in_homogenate,
##           activity_mU_per_mg_protein, within_std_range
## -----------------------------------------------------------------------
calculate_cs_activity <- function(well_rates, protein_by_sample, background_per_sample,
                                   std_slope, std_nmol_max, assay_params, plate_long) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_reps_used = n(), mean_rate_mOD_min = mean(slope_mOD_min),
              sd_rate = sd(slope_mOD_min),
              cv_rate = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop") %>%
    left_join(plate_long %>% filter(well_type == "sample") %>%
                distinct(sample_id, family, individual, temperature),
              by = "sample_id") %>%
    left_join(protein_by_sample %>% select(sample_id, conc_ug_mL, total_protein_mg),
              by = "sample_id") %>%
    left_join(background_per_sample %>% select(sample_id, bg_rate_flat, n_bg_flat),
              by = "sample_id") %>%
    mutate(
      bg_rate_mOD_min       = ifelse(is.na(bg_rate_flat), 0, bg_rate_flat),
      corrected_mOD_min     = mean_rate_mOD_min - bg_rate_mOD_min,
      rate_OD_min           = corrected_mOD_min / 1000,
      nmol_per_min          = rate_OD_min / std_slope,
      activity_mU_per_uL    = nmol_per_min / assay_params$sample_volume_uL *
                              assay_params$dilution_factor,
      total_mU_in_homogenate = activity_mU_per_uL * assay_params$homogenate_volume_uL,
      activity_mU_per_mg_protein = total_mU_in_homogenate / total_protein_mg,
      nmol_in_window        = nmol_per_min * (assay_params$read_duration_min),
      within_std_range      = nmol_in_window <= std_nmol_max
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## build_results_table()
## Purpose : Assemble the final, presentation-ready per-sample results
##           table from cs_activity, cv_summary and baseline diagnostics,
##           with a plain-language Interpretation column.
## Inputs  : cs_activity          - data.frame from calculate_cs_activity()
##           cv_summary           - data.frame from compute_replicate_cv()
##           baseline_per_sample  - data.frame from compute_baseline_diagnostics()
##           assay_params         - list with cv_threshold_pct, homogenate_volume_uL
## Outputs : results_table - formatted data.frame, ready for kable()/write.csv()
## -----------------------------------------------------------------------
build_results_table <- function(cs_activity, cv_summary, baseline_per_sample, assay_params) {
  cs_activity %>%
    left_join(cv_summary %>% select(sample_id, n_all, cv_all), by = "sample_id") %>%
    left_join(baseline_per_sample %>% select(sample_id, n_elevated), by = "sample_id") %>%
    transmute(
      Sample                       = sample_id,
      Family                       = family,
      Individual                   = individual,
      Temperature                  = temperature,
      `Protein conc (ug/mL)`       = round(conc_ug_mL, 1),
      `Total protein (mg)`         = round(total_protein_mg, 3),
      `Reps used`                  = paste0(n_reps_used, "/", n_all),
      `CV all reps (%)`            = round(cv_all, 1),
      `CV used reps (%)`           = round(cv_rate, 1),
      `Rate (mA412/min)`           = round(mean_rate_mOD_min, 2),
      `BG rate (mA412/min)`        = round(bg_rate_mOD_min, 3),
      `Corrected rate (mA412/min)` = round(corrected_mOD_min, 2),
      `Activity (mU/uL)`           = round(activity_mU_per_uL, 4),
      `Activity (mU/mg protein)`   = round(activity_mU_per_mg_protein, 3),
      `CV flag`                    = ifelse(cv_all > assay_params$cv_threshold_pct,
                                            paste0("FAIL >", assay_params$cv_threshold_pct, "%"), "pass"),
      `Elevated baseline reps`     = paste0(n_elevated, "/3"),
      Interpretation               = case_when(
        n_elevated == 3 ~ "DO NOT USE - no clean replicate",
        n_elevated > 0  ~ "caution - some reps compromised",
        cv_rate > assay_params$cv_threshold_pct | is.na(cv_rate) ~ "caution - imprecise",
        TRUE ~ "usable"
      )
    )
}

1.3.5 QC summary and narrative generation

Assembles the per-well QC summary table and auto-generates the plain-language QC findings bullets used in the QC SUMMARY section below.

## -----------------------------------------------------------------------
## build_qc_summary_table()
## Purpose : Consolidate every QC check performed across the pipeline into a
##           single two-column (check, value) table, suitable for kable() and
##           for writing to qc_summary.csv. Mirrors the individual find
##           functions but as compact, at-a-glance figures rather than prose.
## Inputs  : layout_wells, reconcile_summary, std_curve, well_diagnostics,
##           well_rates, cv_summary, baseline_per_sample, pos_control,
##           assay_params
## Outputs : qc_summary - data.frame with columns check, value (both chr)
## -----------------------------------------------------------------------
build_qc_summary_table <- function(layout_wells, reconcile_summary, std_curve,
                                    well_diagnostics, well_rates, cv_summary,
                                    baseline_per_sample, pos_control, assay_params) {
  standard_summary  <- std_curve$standard_summary
  standards_flagged <- std_curve$standards_flagged
  standard_drift    <- std_curve$standard_drift
  sample_rates      <- well_rates %>% filter(well_type == "sample")

  data.frame(
    check = c(
      "Occupied wells in layout",
      "Layout wells missing from absorbance CSV",
      "Max disagreement between absorbance CSV and full report",
      "Standard concentrations with replicate CV > threshold",
      "Standard wells flagged as outliers",
      "Standard curve R^2 (outlier-excluded, replicate level)",
      "Standard curve R^2 (all wells, replicate level)",
      "Standard wells that LOST signal over the run",
      "Standards above photometric linearity ceiling",
      "Background control wells behaving as active reactions",
      "Sample wells with a decreasing trace",
      "Sample wells with an elevated starting A412",
      "Samples with ALL THREE replicates baseline-compromised",
      "Sample wells with a read glitch",
      "Sample wells usable for rate extraction",
      "Samples with technical CV > threshold (all reps)",
      "Samples with technical CV > threshold (usable reps)",
      "Positive control replicates rising and linear"
    ),
    value = c(
      nrow(layout_wells),
      reconcile_summary$n_wells_layout_not_csv,
      sprintf("%.g", reconcile_summary$max_disagreement),
      paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
            collapse = ", "),
      paste0(sum(standards_flagged$is_outlier), "/", nrow(standards_flagged), " (",
             paste(standards_flagged$well[standards_flagged$is_outlier], collapse = ", "), ")"),
      sprintf("%.4f", std_curve$std_r2),
      sprintf("%.4f", summary(std_curve$fit_all_wells)$r.squared),
      paste0(sum(standard_drift$drift < 0), "/", nrow(standard_drift)),
      paste(standard_summary$std_nmol[standard_summary$mean_od > assay_params$od_linear_max],
            collapse = ", "),
      paste0(sum(well_diagnostics$flag_bg_active), "/",
             sum(well_diagnostics$well_type == "background")),
      paste0(sum(well_diagnostics$flag_decreasing & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(well_diagnostics$flag_high_baseline), "/",
             sum(well_diagnostics$well_type == "sample"), " (",
             paste(well_diagnostics$well[well_diagnostics$flag_high_baseline], collapse = ", "), ")"),
      paste0(sum(baseline_per_sample$n_elevated == baseline_per_sample$n), "/",
             nrow(baseline_per_sample), " (",
             paste(baseline_per_sample$sample_id[baseline_per_sample$n_elevated == baseline_per_sample$n],
                   collapse = ", "), ")"),
      paste0(sum(well_diagnostics$flag_glitch & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(sample_rates$rate_usable), "/", nrow(sample_rates)),
      paste0(sum(cv_summary$fails_cv_all), "/", nrow(cv_summary)),
      paste0(sum(cv_summary$fails_cv_usable), "/", nrow(cv_summary)),
      paste0(sum(pos_control$net_change > 0 & pos_control$r2 > 0.99), "/", nrow(pos_control))
    ),
    stringsAsFactors = FALSE
  )
}

## -----------------------------------------------------------------------
## generate_qc_findings()
## Purpose : Walk the QC/CV/anomaly objects for a plate and emit a
##           consistent, generic-language markdown bullet list of findings
##           (decreasing/glitched/high-baseline wells, background-control
##           mix-ups, standard-curve QC failures, CV failures, disqualified
##           samples). Intended to seed the SUMMARY section of a plate's
##           .Rmd; a short hand-written subsection for plate-specific
##           interpretation should follow it.
## Inputs  : well_diagnostics    - data.frame from compute_well_diagnostics()
##           well_rates          - data.frame from compute_well_rates()
##           cv_summary          - data.frame from compute_replicate_cv()
##           baseline_per_sample - data.frame from compute_baseline_diagnostics()
##           std_curve           - list from fit_standard_curve()
##           reconcile_summary   - list from reconcile_kinetic_sources()$summary
##           assay_params        - list with cv_threshold_pct, od_linear_max
## Outputs : character vector, one markdown bullet per finding (empty
##           findings are omitted, so a clean plate returns a short list)
## -----------------------------------------------------------------------
generate_qc_findings <- function(well_diagnostics, well_rates, cv_summary,
                                  baseline_per_sample, std_curve,
                                  reconcile_summary, assay_params) {
  bullets <- character(0)

  # Source reconciliation
  if (reconcile_summary$n_wells_layout_not_csv > 0 ||
      (!is.na(reconcile_summary$max_disagreement) && reconcile_summary$max_disagreement >= 1e-9)) {
    bullets <- c(bullets, sprintf(
      "**Kinetic source disagreement.** %d layout well(s) missing from the absorbance CSV; max disagreement between absorbance CSV and full report was %.4g A412.",
      reconcile_summary$n_wells_layout_not_csv, reconcile_summary$max_disagreement))
  } else {
    bullets <- c(bullets, sprintf(
      "**Kinetic sources agree.** absorbance CSV and full report agree exactly on all %d shared well x timepoint readings; all layout wells are present.",
      reconcile_summary$n_shared_readings))
  }

  # Sample wells with decreasing kinetics
  dec_samples <- well_diagnostics %>% filter(well_type == "sample", flag_decreasing)
  if (nrow(dec_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Decreasing kinetics.** %d sample well(s) end lower than (or equal to) their starting A412 and are excluded from rate extraction: %s.",
      nrow(dec_samples), paste(dec_samples$well, collapse = ", ")))
  }

  # Read glitches
  glitch_samples <- well_diagnostics %>% filter(well_type == "sample", flag_glitch)
  if (nrow(glitch_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Read glitches.** %d sample well(s) contain a single read-to-read step inconsistent with its neighbours: %s (only disqualifying if it falls inside the fitted rate window).",
      nrow(glitch_samples), paste(glitch_samples$well, collapse = ", ")))
  }

  # Elevated baseline
  high_baseline <- well_diagnostics %>% filter(well_type == "sample", flag_high_baseline)
  if (nrow(high_baseline) > 0) {
    bullets <- c(bullets, sprintf(
      "**Elevated starting absorbance.** %d sample well(s) start above the baseline threshold (A412 > %s), indicating pre-existing thiol/contamination before the reaction began: %s.",
      nrow(high_baseline), assay_params$sample_baseline_od, paste(high_baseline$well, collapse = ", ")))
  }

  # Samples fully disqualified by baseline
  disqualified <- baseline_per_sample %>% filter(n_elevated == n)
  if (nrow(disqualified) > 0) {
    bullets <- c(bullets, sprintf(
      "**Sample(s) with no clean replicate.** Every replicate of %s has an elevated starting A412; these samples have no usable measurement on this plate regardless of CV.",
      paste(disqualified$sample_id, collapse = ", ")))
  }

  # Over-range wells
  over_range <- well_diagnostics %>% filter(flag_over_range)
  if (nrow(over_range) > 0) {
    bullets <- c(bullets, sprintf(
      "**Over-range wells.** %d well(s) exceed the photometric linearity ceiling (A412 > %s): %s.",
      nrow(over_range), assay_params$od_linear_max, paste(over_range$well, collapse = ", ")))
  }

  # Background wells behaving as active reactions
  bg_active <- well_diagnostics %>% filter(well_type == "background", flag_bg_active)
  if (nrow(bg_active) > 0) {
    bullets <- c(bullets, sprintf(
      "**Background control wells behaving as active reactions.** %d of %d background wells show a starting A412 or drift inconsistent with a flat, inactive well: %s.",
      nrow(bg_active), sum(well_diagnostics$well_type == "background"),
      paste(bg_active$well, collapse = ", ")))
  }

  # Standard curve QC
  std_summary <- std_curve$standard_summary
  cv_fail_std <- std_summary %>% filter(cv_pct > assay_params$cv_threshold_pct)
  if (nrow(cv_fail_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**GSH standard curve replicate imprecision.** %d of %d concentrations exceed %s%% CV: %s nmol/well.",
      nrow(cv_fail_std), nrow(std_summary), assay_params$cv_threshold_pct,
      paste(cv_fail_std$std_nmol, collapse = ", ")))
  }
  n_std_outliers <- sum(std_curve$standards_flagged$is_outlier)
  if (n_std_outliers > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard outlier wells.** %d of %d standard wells deviate from their triplicate median by more than %s A412: %s.",
      n_std_outliers, nrow(std_curve$standards_flagged), assay_params$std_outlier_od,
      paste(std_curve$standards_flagged$well[std_curve$standards_flagged$is_outlier], collapse = ", ")))
  }
  over_range_std <- std_summary %>% filter(mean_od > assay_params$od_linear_max)
  if (nrow(over_range_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**Standards above the photometric linear range.** %s nmol/well read above A412 %s; the calibration is trustworthy only below these concentrations.",
      paste(over_range_std$std_nmol, collapse = ", "), assay_params$od_linear_max))
  }
  n_drift_neg <- sum(std_curve$standard_drift$drift < 0)
  if (n_drift_neg > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard signal decay.** %d of %d standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.",
      n_drift_neg, nrow(std_curve$standard_drift)))
  }
  bullets <- c(bullets, sprintf(
    "**Standard curve fit (outlier-excluded):** slope = %.5f A412/nmol, R^2 = %.4f (all-wells R^2 = %.4f).",
    std_curve$std_slope, std_curve$std_r2, summary(std_curve$fit_all_wells)$r.squared))

  # Technical replicate CV failures
  fail_all <- cv_summary %>% filter(fails_cv_all)
  if (nrow(fail_all) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% (all replicates).** %d of %d samples: %s.",
      assay_params$cv_threshold_pct, nrow(fail_all), nrow(cv_summary),
      paste(sprintf("%s (%.0f%%)", fail_all$sample_id, fail_all$cv_all), collapse = ", ")))
  } else {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV.** All %d samples are within %s%% CV across all three replicates.",
      nrow(cv_summary), assay_params$cv_threshold_pct))
  }
  fail_usable <- cv_summary %>% filter(fails_cv_usable)
  if (nrow(fail_usable) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% persists after excluding flagged replicates.** %s.",
      assay_params$cv_threshold_pct,
      paste(sprintf("%s (%.0f%%, n=%d)", fail_usable$sample_id, fail_usable$cv_usable,
                    fail_usable$n_usable), collapse = ", ")))
  }

  bullets
}

1.4 Assay parameters

All assay constants and QC thresholds are collected here so nothing downstream hard-codes a number.

assay_params <- list(
  sample_volume_uL      = 2,     # homogenate volume per reaction well (V in Abcam formula)
  buffer_volume_uL      = 48,    # Assay Buffer 7 added to reach 50 uL pre-Reaction Mix
  dilution_factor       = 1,     # D; layout `df.0` = undiluted homogenate
  homogenate_volume_uL  = 350,   # MEASURED buffer volume tissue was homogenized in (see note 2)
  read_wavelength_nm    = 412,
  read_interval_min     = 2,
  read_duration_min     = 20,    # see note 1
  rate_window_n         = 5,     # points per sliding regression window (5 pts = 8 min)
  rate_min_r2           = 0.80,  # min R^2 for a rate window to be trusted
  glitch_excess_od      = 0.02,  # |step - mean of neighbouring steps| flagging a read glitch
  cv_threshold_pct      = 15,    # technical-replicate CV QC threshold
  od_linear_max         = 1.5,   # upper A412 bound for reliable photometry
  std_outlier_od        = 0.15,  # |deviation from triplicate median| flagging a standard well
  sample_baseline_od    = 0.35,  # max acceptable t0 A412 for a sample well
  bg_flat_od_max        = 0.15,  # a well-behaved background well starts below this A412
  bg_flat_drift_max     = 0.05   # ...and drifts less than this over the whole run
)

cat("--- assay_params: assay constants and QC thresholds ---\n\n")
str(assay_params)
--- assay_params: assay constants and QC thresholds ---

List of 16
 $ sample_volume_uL    : num 2
 $ buffer_volume_uL    : num 48
 $ dilution_factor     : num 1
 $ homogenate_volume_uL: num 350
 $ read_wavelength_nm  : num 412
 $ read_interval_min   : num 2
 $ read_duration_min   : num 20
 $ rate_window_n       : num 5
 $ rate_min_r2         : num 0.8
 $ glitch_excess_od    : num 0.02
 $ cv_threshold_pct    : num 15
 $ od_linear_max       : num 1.5
 $ std_outlier_od      : num 0.15
 $ sample_baseline_od  : num 0.35
 $ bg_flat_od_max      : num 0.15
 $ bg_flat_drift_max   : num 0.05

1.5 Output directory

# Output directory for this analysis (matches this file's name, per ../code/README.md)
output_dir <- "../outputs/Gen5-20260814-mgig-citrate_synthase-F05-36C"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- output_dir: destination for all figures and tables ---\n")
str(output_dir)
--- output_dir: destination for all figures and tables ---
 chr "../outputs/Gen5-20260814-mgig-citrate_synthase-F05-36C"

2 DATA

Data are read from the local repo (../data/raw_absorbance/) so this document renders before/after the files are pushed to GitHub.

2.1 Plate layout

data_dir <- "../data/raw_absorbance"
run_stem <- "Gen5-20260814-mgig-citrate_synthase-F05-36C"

layout_wells <- parse_plate_layout(file.path(data_dir, paste0("layout-", run_stem, ".csv")))

cat("Layout format detected:", attr(layout_wells, "layout_format"), "\n")
cat("Occupied wells in layout:", nrow(layout_wells), "\n")

cat("\n--- layout_wells: one row per occupied well, with its descriptive label ---\n\n")
str(layout_wells)
Layout format detected: single 
Occupied wells in layout: 69 

--- layout_wells: one row per occupied well, with its descriptive label ---

'data.frame':   69 obs. of  4 variables:
 $ well     : chr  "A1" "A2" "A3" "A4" ...
 $ plate_row: chr  "A" "A" "A" "A" ...
 $ plate_col: int  1 2 3 4 5 6 7 8 9 10 ...
 $ label    : chr  "F05_01_36C-citrate_synthase-12.8-df.0" "F05_01_36C-citrate_synthase-12.8-df.0" "F05_01_36C-citrate_synthase-12.8-df.0" "F05_02_36C-citrate_synthase-6.2-df.0" ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "row_idx" "col_idx"
  ..$ dimnames:List of 2
  .. ..$ row_idx: chr [1:8] "row_idx=2" "row_idx=3" "row_idx=4" "row_idx=5" ...
  .. ..$ col_idx: chr [1:12] "col_idx= 2" "col_idx= 3" "col_idx= 4" "col_idx= 5" ...
 - attr(*, "layout_format")= chr "single"

2.2 Kinetic readers

absorbance_csv <- read_absorbance_csv(file.path(data_dir, paste0("absorbance-", run_stem, ".csv")))
full_report    <- read_full_report(file.path(data_dir, paste0("full_report-", run_stem, ".txt")))

cat("absorbance CSV : ", dplyr::n_distinct(absorbance_csv$well), " wells x ",
    dplyr::n_distinct(absorbance_csv$time_min), " timepoints\n", sep = "")
cat("full report    : ", dplyr::n_distinct(full_report$well), " wells x ",
    dplyr::n_distinct(full_report$time_min), " timepoints\n", sep = "")

cat("\n--- absorbance_csv: long-format readings from absorbance-*.csv ---\n\n")
str(absorbance_csv)
cat("\n--- full_report: long-format readings from full_report-*.txt ---\n\n")
str(full_report)
absorbance CSV : 69 wells x 11 timepoints
full report    : 69 wells x 11 timepoints

--- absorbance_csv: long-format readings from absorbance-*.csv ---

'data.frame':   759 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.179 0.182 0.186 0.19 0.194 0.199 0.203 0.207 0.212 0.217 ...

--- full_report: long-format readings from full_report-*.txt ---

'data.frame':   759 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.179 0.182 0.186 0.19 0.194 0.199 0.203 0.207 0.212 0.217 ...

2.3 Cross-check against the full report

absorbance-*.csv is used for every well; full_report-*.txt is used only to confirm the CSV is trustworthy: every well it reports should also be in the CSV, every well the layout expects should be in the CSV, and every reading the two files share should agree exactly.

recon <- reconcile_kinetic_sources(absorbance_csv, full_report, layout_wells)
plate_readings <- recon$plate_readings

cat("Wells in absorbance CSV          :", recon$summary$n_wells_csv, "\n")
cat("Wells in full report             :", recon$summary$n_wells_report, "\n")
cat("Wells in full report not in CSV  :", recon$summary$n_wells_report_not_csv, "\n")
cat("Layout wells missing from CSV    :", recon$summary$n_wells_layout_not_csv, "\n")
cat("CSV wells not in layout          :", recon$summary$n_wells_csv_not_layout, "\n")
cat("Shared well x timepoints         :", recon$summary$n_shared_readings, "\n")
cat("Max |CSV - report| disagreement  :", recon$summary$max_disagreement, "\n")

cat("\n--- recon$overlap_check: per-reading comparison of the two raw files ---\n\n")
str(recon$overlap_check)
cat("\n--- plate_readings: readings for all occupied wells, from absorbance_csv ---\n\n")
str(plate_readings)
Wells in absorbance CSV          : 69 
Wells in full report             : 69 
Wells in full report not in CSV  : 0 
Layout wells missing from CSV    : 0 
CSV wells not in layout          : 0 
Shared well x timepoints         : 759 
Max |CSV - report| disagreement  : 0 

--- recon$overlap_check: per-reading comparison of the two raw files ---

'data.frame':   759 obs. of  5 variables:
 $ well     : chr  "A1" "A1" "A1" "A1" ...
 $ time_min : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od_csv   : num  0.179 0.182 0.186 0.19 0.194 0.199 0.203 0.207 0.212 0.217 ...
 $ od_report: num  0.179 0.182 0.186 0.19 0.194 0.199 0.203 0.207 0.212 0.217 ...
 $ abs_diff : num  0 0 0 0 0 0 0 0 0 0 ...

--- plate_readings: readings for all occupied wells, from absorbance_csv ---

'data.frame':   759 obs. of  4 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.179 0.182 0.186 0.19 0.194 0.199 0.203 0.207 0.212 0.217 ...
 $ source  : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...

This plate’s absorbance-*.csv covers the full 69 occupied wells and the layout carries no spelling typo, so no label fix is needed here.

2.4 Annotate wells and parse metadata

plate_long <- suppressWarnings(annotate_wells(plate_readings, layout_wells))

cat("Wells per type:\n")
print(plate_long %>% distinct(well, well_type) %>% count(well_type, name = "n_wells") %>%
        as.data.frame(), row.names = FALSE)
cat("\nSamples:", paste(sort(unique(na.omit(plate_long$sample_id))), collapse = ", "), "\n")
cat("Standards (nmol/well):", paste(sort(unique(na.omit(plate_long$std_nmol))), collapse = ", "), "\n")

cat("\n--- plate_long: fully annotated long-format plate, one row per well x timepoint ---\n\n")
str(plate_long)
Wells per type:
        well_type n_wells
       background      24
 positive_control       3
           sample      24
         standard      18

Samples: F05_01_36C, F05_02_36C, F05_03_36C, F05_04_36C, F05_05_36C, F05_06_36C, F05_07_36C, F05_08_36C 
Standards (nmol/well): 0, 8, 16, 24, 32, 40 

--- plate_long: fully annotated long-format plate, one row per well x timepoint ---

'data.frame':   759 obs. of  15 variables:
 $ well       : chr  "C1" "C1" "C1" "C1" ...
 $ time_min   : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od         : num  0.047 0.047 0.047 0.048 0.048 0.048 0.048 0.048 0.048 0.048 ...
 $ source     : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ plate_row  : chr  "C" "C" "C" "C" ...
 $ plate_col  : int  1 1 1 1 1 1 1 1 1 1 ...
 $ label      : chr  "F05_01_36C_BG-citrate_synthase-12.8-df.0" "F05_01_36C_BG-citrate_synthase-12.8-df.0" "F05_01_36C_BG-citrate_synthase-12.8-df.0" "F05_01_36C_BG-citrate_synthase-12.8-df.0" ...
 $ well_type  : chr  "background" "background" "background" "background" ...
 $ label_clean: chr  "F05_01_36C_BG-citrate_synthase-12.8-df.0" "F05_01_36C_BG-citrate_synthase-12.8-df.0" "F05_01_36C_BG-citrate_synthase-12.8-df.0" "F05_01_36C_BG-citrate_synthase-12.8-df.0" ...
 $ sample_id  : chr  "F05_01_36C" "F05_01_36C" "F05_01_36C" "F05_01_36C" ...
 $ family     : chr  "F05" "F05" "F05" "F05" ...
 $ individual : chr  "01" "01" "01" "01" ...
 $ temperature: chr  "36C" "36C" "36C" "36C" ...
 $ weight_mg  : num  12.8 12.8 12.8 12.8 12.8 12.8 12.8 12.8 12.8 12.8 ...
 $ std_nmol   : num  NA NA NA NA NA NA NA NA NA NA ...

annotate_wells() emits a harmless NAs introduced by coercion warning (suppressed above), because its weight_mg parser only applies to sample/background wells and returns NA for standards/positive control by construction. No weight_mg value is actually missing for a sample or background well; annotate_wells()’s internal stopifnot() would halt otherwise.

2.5 Protein concentration (normalization factor)

Activity is normalized to total extracted protein rather than tissue weight (note 2). Protein concentration for each sample was measured on a separate BCA/Bradford-style plate (595 nm) and supplied as two CSV exports covering different subsets of samples across families F05 and F07. This plate’s F05_07_36C record appears only in file 1, under the string F-05_07_36C (stray hyphen); it is corrected with sample_id_fixes before matching.

protein_files <- c(
  "../data/BSA/raw_absorbance/sample_protein_concentrations_1.csv",
  "../data/BSA/raw_absorbance/sample_protein_concentrations_2.csv"
)

plate_sample_ids <- sort(unique(na.omit(plate_long$sample_id)))

protein_by_sample <- load_protein_concentrations(protein_files, plate_sample_ids,
                                                  assay_params$homogenate_volume_uL,
                                                  sample_id_fixes = c("F-05_07_36C" = "F05_07_36C")) %>%
  left_join(plate_long %>% filter(well_type == "sample") %>%
              distinct(sample_id, family, individual, temperature),
            by = "sample_id")

cat("Samples matched to a protein concentration record:", nrow(protein_by_sample),
    "/", length(plate_sample_ids), "\n")
cat("Protein concentration range (ug/mL):",
    paste(range(protein_by_sample$conc_ug_mL), collapse = " - "), "\n")
cat("Total protein per homogenate range (mg):",
    paste(round(range(protein_by_sample$total_protein_mg), 3), collapse = " - "), "\n\n")
print(protein_by_sample %>% select(sample_id, source_file, conc_ug_mL, total_protein_mg) %>%
        as.data.frame(), row.names = FALSE)

cat("\n--- protein_by_sample: matched protein concentration and total protein per homogenate ---\n\n")
str(protein_by_sample)
Samples matched to a protein concentration record: 8 / 8 
Protein concentration range (ug/mL): 545.8 - 1666.8 
Total protein per homogenate range (mg): 0.191 - 0.583 

  sample_id                         source_file conc_ug_mL total_protein_mg
 F05_01_36C sample_protein_concentrations_1.csv      953.1         0.333585
 F05_02_36C sample_protein_concentrations_1.csv      545.8         0.191030
 F05_03_36C sample_protein_concentrations_1.csv      667.6         0.233660
 F05_04_36C sample_protein_concentrations_1.csv     1062.2         0.371770
 F05_05_36C sample_protein_concentrations_1.csv     1666.8         0.583380
 F05_06_36C sample_protein_concentrations_1.csv     1265.9         0.443065
 F05_07_36C sample_protein_concentrations_1.csv     1110.5         0.388675
 F05_08_36C sample_protein_concentrations_1.csv     1357.5         0.475125

--- protein_by_sample: matched protein concentration and total protein per homogenate ---

'data.frame':   8 obs. of  7 variables:
 $ sample_id       : chr  "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ source_file     : chr  "sample_protein_concentrations_1.csv" "sample_protein_concentrations_1.csv" "sample_protein_concentrations_1.csv" "sample_protein_concentrations_1.csv" ...
 $ conc_ug_mL      : num  953 546 668 1062 1667 ...
 $ total_protein_mg: num  0.334 0.191 0.234 0.372 0.583 ...
 $ family          : chr  "F05" "F05" "F05" "F05" ...
 $ individual      : chr  "01" "02" "03" "04" ...
 $ temperature     : chr  "36C" "36C" "36C" "36C" ...

3 KINETIC TRACES

Every trace is plotted before any rate is extracted. In this assay the shape of the trace is the primary diagnostic: CS activity must produce a monotonically rising A412, and only a genuinely linear stretch may be used for a rate.

trace_plot <- ggplot(plate_long, aes(x = time_min, y = od, group = well, colour = well_type)) +
  geom_line(linewidth = 0.5) +
  geom_hline(yintercept = assay_params$od_linear_max,
             linetype = "dotted", colour = "grey40") +
  facet_wrap(~ plate_row + plate_col, ncol = 12,
             labeller = function(d) list(paste0(d$plate_row, d$plate_col))) +
  scale_colour_manual(values = c(sample = "#08519c", background = "#a63603",
                                 standard = "#238b45", positive_control = "#6a51a3")) +
  labs(title = "A412 kinetic traces, all occupied wells",
       subtitle = paste0("Dotted line = assumed photometric linearity ceiling (A412 = ",
                         assay_params$od_linear_max,
                         "). Rising traces only are usable for rate extraction."),
       x = "Time (min)", y = "A412", colour = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        strip.text = element_text(size = 6),
        axis.text  = element_text(size = 5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "kinetic_traces_all_wells.png"), trace_plot,
       width = 11, height = 8, dpi = 300)

cat("--- trace_plot: ggplot object structure ---\n\n")
summary(trace_plot)

trace_plot

--- trace_plot: ggplot object structure ---

data: well, time_min, od, source, plate_row, plate_col, label,
  well_type, label_clean, sample_id, family, individual, temperature,
  weight_mg, std_nmol [759x15]
mapping:  x = ~time_min, y = ~od, group = ~well, colour = ~well_type
scales:   colour 
faceting:  ~plate_row, ~plate_col 
-----------------------------------
geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

Every sample and positive-control trace on this plate rises smoothly and close to linearly across the full 20-minute window (replicate-level R2 > 0.99 for a full-trace linear fit on every sample well, computed below). This plate shows no elevated-starting-absorbance or contamination signature in any well.

4 ANOMALY DETECTION

4.1 Per-well trace diagnostics

Four independent checks, computed from the raw trace before any rate fitting:

  1. Direction — net A412 change from first to last read. A CS-containing well must be positive.
  2. Monotonicity — fraction of the 10 read-to-read intervals that rise.
  3. Over-range — any read above the photometric linearity ceiling.
  4. Discontinuity — a single read-to-read step that departs from its two neighbouring steps, catching read glitches rather than biology.

Two further checks specific to well role: an elevated starting A412 (pre-existing thiol/contamination) for sample wells, and a background well whose trace is not flat (indicating it received Reaction Mix rather than Background Control Mix).

well_diagnostics <- compute_well_diagnostics(plate_long, assay_params)

cat("Flag counts across all", nrow(well_diagnostics), "wells:\n")
print(well_diagnostics %>% summarise(across(starts_with("flag_"), sum)) %>%
        as.data.frame(), row.names = FALSE)

cat("\nWells with at least one flag:", sum(well_diagnostics$n_flags > 0), "/", nrow(well_diagnostics), "\n")

cat("\n--- well_diagnostics: per-well trace diagnostics and anomaly flags ---\n\n")
str(well_diagnostics)
Flag counts across all 69 wells:
 flag_decreasing flag_over_range flag_glitch flag_high_baseline flag_bg_active
               4               0           1                  0              1

Wells with at least one flag: 5 / 69 

--- well_diagnostics: per-well trace diagnostics and anomaly flags ---

tibble [69 × 24] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:69] "A1" "A10" "A11" "A12" ...
 $ plate_row         : chr [1:69] "A" "A" "A" "A" ...
 $ plate_col         : int [1:69] 1 10 11 12 2 3 4 5 6 7 ...
 $ well_type         : chr [1:69] "sample" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:69] "F05_01_36C" "F05_04_36C" "F05_04_36C" "F05_04_36C" ...
 $ label             : chr [1:69] "F05_01_36C-citrate_synthase-12.8-df.0" "F05_04_36C-citrate_synthase-12.8-df.0" "F05_04_36C-citrate_synthase-12.8-df.0" "F05_04_36C-citrate_synthase-12.8-df.0" ...
 $ std_nmol          : num [1:69] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:69] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:69] 0.179 0.185 0.183 0.183 0.182 0.18 0.184 0.183 0.182 0.182 ...
 $ od_last           : num [1:69] 0.222 0.228 0.227 0.228 0.224 0.221 0.226 0.226 0.224 0.227 ...
 $ od_max            : num [1:69] 0.222 0.228 0.227 0.228 0.224 0.221 0.226 0.226 0.224 0.227 ...
 $ net_change        : num [1:69] 0.043 0.043 0.044 0.045 0.042 0.041 0.042 0.043 0.042 0.045 ...
 $ frac_rising       : num [1:69] 1 1 1 1 1 1 1 1 1 1 ...
 $ max_step          : num [1:69] 0.005 0.005 0.005 0.005 0.005 0.005 0.005 0.005 0.005 0.005 ...
 $ typical_step      : num [1:69] 0.004 0.0045 0.0045 0.0045 0.004 ...
 $ step_excess       : num [1:69] -0.001 0.0015 -0.001 -0.0005 0.001 ...
 $ glitch_at_min     : num [1:69] 2 8 12 10 14 16 2 10 6 2 ...
 $ step_ratio        : num [1:69] 1.25 1.11 1.11 1.11 1.25 ...
 $ flag_decreasing   : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range   : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:69] 0 0 0 0 0 0 0 0 0 0 ...

4.2 Anomaly table

anomaly_table <- well_diagnostics %>%
  filter(n_flags > 0) %>%
  select(well, sample_id, well_type, od_first, net_change, n_flags,
         flag_decreasing, flag_over_range, flag_glitch,
         flag_high_baseline, flag_bg_active) %>%
  arrange(desc(n_flags), well)

kable(anomaly_table, digits = 3,
      caption = "Wells with at least one anomaly flag")

cat("\nTotal flagged wells:", nrow(anomaly_table), "/", nrow(well_diagnostics), "\n")

cat("\n--- anomaly_table: all flagged wells with their specific flags ---\n\n")
str(anomaly_table)
well sample_id well_type od_first net_change n_flags flag_decreasing flag_over_range flag_glitch flag_high_baseline flag_bg_active
C8 F05_03_36C background 0.252 -0.050 2 TRUE FALSE FALSE FALSE TRUE
F1 NA standard 1.144 -0.004 1 TRUE FALSE FALSE FALSE FALSE
F4 NA standard 1.433 -0.008 1 TRUE FALSE FALSE FALSE FALSE
F5 NA standard 1.067 0.079 1 FALSE FALSE TRUE FALSE FALSE
F6 NA standard 1.422 -0.011 1 TRUE FALSE FALSE FALSE FALSE

Wells with at least one anomaly flag

Total flagged wells: 5 / 69 

--- anomaly_table: all flagged wells with their specific flags ---

tibble [5 × 11] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:5] "C8" "F1" "F4" "F5" ...
 $ sample_id         : chr [1:5] "F05_03_36C" NA NA NA ...
 $ well_type         : chr [1:5] "background" "standard" "standard" "standard" ...
 $ od_first          : num [1:5] 0.252 1.144 1.433 1.067 1.422
 $ net_change        : num [1:5] -0.05 -0.004 -0.008 0.079 -0.011
 $ n_flags           : int [1:5] 2 1 1 1 1
 $ flag_decreasing   : logi [1:5] TRUE TRUE TRUE FALSE TRUE
 $ flag_over_range   : logi [1:5] FALSE FALSE FALSE FALSE FALSE
 $ flag_glitch       : logi [1:5] FALSE FALSE FALSE TRUE FALSE
 $ flag_high_baseline: logi [1:5] FALSE FALSE FALSE FALSE FALSE
 $ flag_bg_active    : logi [1:5] TRUE FALSE FALSE FALSE FALSE

Only two wells are flagged on this entire plate, both isolated, single-flag cases with no counterpart pattern elsewhere on the plate:

  • C8 (background, F05_03_36C) is flagged flag_bg_active and flag_decreasing together: it starts at A412 0.252 (above the bg_flat_od_max of 0.15) and then declines steadily to 0.202 (net change -0.050), rather than staying flat like a true background well. This pattern is consistent with Reaction Mix having been dispensed in place of Background Control Mix for this single well.
  • F5 (the 40 nmol GSH standard) is flagged flag_glitch: see the GSH STANDARD CURVE section below, where it is also the sole outlier by the std_outlier_od triplicate-deviation check.

No sample well is flagged for anything on this plate.

4.3 Background wells behaving as active reactions

bg_active_wells <- well_diagnostics %>%
  filter(well_type == "background", flag_bg_active) %>%
  select(well, sample_id, od_first, net_change)

kable(bg_active_wells, digits = 3,
      caption = "Background wells with a non-flat trace")

cat("Background wells behaving as active reactions:", nrow(bg_active_wells),
    "/", sum(well_diagnostics$well_type == "background"), "\n")

cat("\n--- bg_active_wells: background wells excluded from the background-rate estimate ---\n\n")
str(bg_active_wells)
well sample_id od_first net_change
C8 F05_03_36C 0.252 -0.05

Background wells with a non-flat trace

Background wells behaving as active reactions: 1 / 24 

--- bg_active_wells: background wells excluded from the background-rate estimate ---

tibble [1 × 4] (S3: tbl_df/tbl/data.frame)
 $ well      : chr "C8"
 $ sample_id : chr "F05_03_36C"
 $ od_first  : num 0.252
 $ net_change: num -0.05

Only C8 is affected — 1 of 24 background wells. Because compute_background_correction() estimates the background rate from only the flat replicates within each sample’s background triplicate, C8 simply drops out of the F05_03_36C background estimate (2 of 3 remain), and every other sample’s background estimate is unaffected.

4.4 Baseline diagnostic (elevated starting absorbance)

bd <- compute_baseline_diagnostics(well_diagnostics, assay_params)
baseline_check       <- bd$baseline_check
baseline_per_sample  <- bd$baseline_per_sample

cat("Sample/background wells with elevated starting A412 (> ",
    assay_params$sample_baseline_od, "):",
    sum(baseline_check$baseline == "elevated"), "/", nrow(baseline_check), "\n", sep = "")

kable(baseline_per_sample, digits = 3,
      caption = "Elevated-baseline replicate count per sample")

cat("\n--- baseline_check: per-well baseline classification ---\n\n")
str(baseline_check)
cat("\n--- baseline_per_sample: elevated-replicate count per sample ---\n\n")
str(baseline_per_sample)
Sample/background wells with elevated starting A412 (> 0.35):0/48
sample_id n_elevated n median_baseline
F05_01_36C 0 3 0.180
F05_02_36C 0 3 0.183
F05_03_36C 0 3 0.182
F05_04_36C 0 3 0.183
F05_05_36C 0 3 0.190
F05_06_36C 0 3 0.183
F05_07_36C 0 3 0.184
F05_08_36C 0 3 0.185

Elevated-baseline replicate count per sample

--- baseline_check: per-well baseline classification ---

tibble [48 × 25] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:48] "C8" "B1" "B2" "B3" ...
 $ plate_row         : chr [1:48] "C" "B" "B" "B" ...
 $ plate_col         : int [1:48] 8 1 2 3 5 11 10 12 7 4 ...
 $ well_type         : chr [1:48] "background" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:48] "F05_03_36C" "F05_05_36C" "F05_05_36C" "F05_05_36C" ...
 $ label             : chr [1:48] "F05_03_36C_BG-citrate_synthase-10.3-df.0" "F05_05_36C-citrate_synthase-22.7-df.0" "F05_05_36C-citrate_synthase-22.7-df.0" "F05_05_36C-citrate_synthase-22.7-df.0" ...
 $ std_nmol          : num [1:48] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:48] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:48] 0.252 0.192 0.19 0.19 0.188 0.186 0.185 0.185 0.185 0.184 ...
 $ od_last           : num [1:48] 0.202 0.235 0.231 0.231 0.23 0.227 0.228 0.23 0.227 0.226 ...
 $ od_max            : num [1:48] 0.254 0.235 0.231 0.231 0.23 0.227 0.228 0.23 0.227 0.226 ...
 $ net_change        : num [1:48] -0.05 0.043 0.041 0.041 0.042 0.041 0.043 0.045 0.042 0.042 ...
 $ frac_rising       : num [1:48] 0.1 1 1 1 1 1 1 1 1 1 ...
 $ max_step          : num [1:48] -0.008 0.005 0.005 0.005 0.005 0.005 0.005 0.006 0.005 0.005 ...
 $ typical_step      : num [1:48] 0.0055 0.004 0.004 0.004 0.004 ...
 $ step_excess       : num [1:48] 0.007 -0.0015 0.001 -0.001 -0.0015 ...
 $ glitch_at_min     : num [1:48] 2 6 2 2 8 14 8 2 2 2 ...
 $ step_ratio        : num [1:48] 1.45 1.25 1.25 1.25 1.25 ...
 $ flag_decreasing   : logi [1:48] TRUE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range   : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:48] TRUE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:48] 2 0 0 0 0 0 0 0 0 0 ...
 $ baseline          : chr [1:48] "normal" "normal" "normal" "normal" ...

--- baseline_per_sample: elevated-replicate count per sample ---

tibble [8 × 4] (S3: tbl_df/tbl/data.frame)
 $ sample_id      : chr [1:8] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ n_elevated     : int [1:8] 0 0 0 0 0 0 0 0
 $ n              : int [1:8] 3 3 3 3 3 3 3 3
 $ median_baseline: num [1:8] 0.18 0.183 0.182 0.183 0.19 0.183 0.184 0.185

No sample or background well anywhere on this plate starts above A412 0.35 — sample wells all start within a tight 0.179-0.192 range (well below the 0.35 threshold), and no sample has any elevated-baseline replicate. No sample is disqualified on this basis.

4.5 Plot starting absorbance

baseline_plot <- ggplot(baseline_check, aes(x = reorder(well, od_first), y = od_first,
                                            fill = well_type)) +
  geom_col() +
  geom_hline(yintercept = assay_params$sample_baseline_od,
             linetype = "dashed", colour = "#cb181d") +
  scale_fill_manual(values = c(sample = "#08519c", background = "#a63603")) +
  labs(title = "Starting A412 (t = 0) for every sample/background well",
       subtitle = paste0("Dashed line = elevated-baseline threshold (A412 = ",
                         assay_params$sample_baseline_od, "). No well on this plate exceeds it."),
       x = "Well", y = "A412 at t = 0", fill = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 90, size = 5, vjust = 0.5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "starting_absorbance.png"), baseline_plot,
       width = 9, height = 6, dpi = 300)

cat("--- baseline_plot: ggplot object structure ---\n\n")
summary(baseline_plot)

baseline_plot

--- baseline_plot: ggplot object structure ---

data: well, plate_row, plate_col, well_type, sample_id, label,
  std_nmol, source, od_first, od_last, od_max, net_change, frac_rising,
  max_step, typical_step, step_excess, glitch_at_min, step_ratio,
  flag_decreasing, flag_over_range, flag_glitch, flag_high_baseline,
  flag_bg_active, n_flags, baseline [48x25]
mapping:  x = ~reorder(well, od_first), y = ~od_first, fill = ~well_type
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

5 GSH STANDARD CURVE

The GSH standard is a stoichiometric endpoint, not an enzymatic reaction: free thiol reduces DTNB immediately on mixing. The standard curve is therefore read from a single timepoint, and t = 0 is used because TNB2- degrades over the run (quantified below).

5.1 Fit the standard curve

Three candidate fits are compared:

  • all wells — every standard replicate, no exclusions
  • concentration means — the conventional fit, on per-concentration means
  • outlier-excluded — replicate-level fit after dropping flagged wells

The outlier-excluded fit is used for activity calculation.

std_curve <- fit_standard_curve(plate_long, assay_params)

standards_t0       <- std_curve$standards_t0
standard_drift     <- std_curve$standard_drift
standard_summary   <- std_curve$standard_summary
standards_flagged  <- std_curve$standards_flagged
fit_all_wells      <- std_curve$fit_all_wells
fit_comparison     <- std_curve$fit_comparison
std_slope          <- std_curve$std_slope
std_intercept      <- std_curve$std_intercept
std_r2             <- std_curve$std_r2
std_nmol_max       <- std_curve$std_nmol_max

cat("Standard wells:", nrow(standards_t0), "at", dplyr::n_distinct(standards_t0$std_nmol),
    "concentrations\n")
cat("Standard drift over 20 min (dA412): median",
    sprintf("%.3f", median(standard_drift$drift)),
    "| range", paste(sprintf("%.3f", range(standard_drift$drift)), collapse = " to "), "\n")
cat("Standards that LOST signal (drift < 0):", sum(standard_drift$drift < 0), "/", nrow(standard_drift), "\n")

cat("\n--- standard_summary: per-concentration mean/SD/CV of t0 absorbance ---\n\n")
str(standard_summary)
kable(standard_summary, digits = c(0, 0, 3, 4, 4, 2, 3, 3, 4),
      caption = "GSH standard curve, per-concentration summary at t = 0")

cat("\nConcentrations exceeding", assay_params$cv_threshold_pct, "% replicate CV:",
    paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
          collapse = ", "), "\n")

cat("\n--- standards_flagged: every standard well with its deviation from the triplicate median ---\n\n")
str(standards_flagged)
kable(standards_flagged %>% filter(is_outlier) %>% select(well, std_nmol, od, triplicate_median, deviation),
      digits = 3, caption = "Standard wells flagged as outliers (|deviation| > 0.15 A412)")

cat("\n--- fit_comparison: slope/intercept/R^2 for all three candidate fits ---\n\n")
str(fit_comparison)
# row.names = FALSE: coef() attaches the predictor's name ("std_nmol") to the
# slope/intercept elements, which data.frame() otherwise promotes to a mangled
# row label (e.g. "std_nmol...1") that breaks pandoc's pipe-table parser.
kable(fit_comparison, digits = c(0, 0, 5, 4, 5), row.names = FALSE,
      caption = "Standard curve fit comparison (outlier-excluded fit used downstream)")
Standard wells: 18 at 6 concentrations
Standard drift over 20 min (dA412): median 0.030 | range -0.011 to 0.079 
Standards that LOST signal (drift < 0): 3 / 18 

--- standard_summary: per-concentration mean/SD/CV of t0 absorbance ---

tibble [6 × 9] (S3: tbl_df/tbl/data.frame)
 $ std_nmol   : num [1:6] 0 8 16 24 32 40
 $ n          : int [1:6] 3 3 3 3 3 3
 $ mean_od    : num [1:6] 0.17 0.448 0.692 0.918 1.153 ...
 $ sd_od      : num [1:6] 0.00802 0.00586 0.006 0.00929 0.01097 ...
 $ se_od      : num [1:6] 0.00463 0.00338 0.00346 0.00536 0.00633 ...
 $ cv_pct     : num [1:6] 4.727 1.309 0.867 1.012 0.952 ...
 $ median_od  : num [1:6] 0.169 0.45 0.692 0.921 1.149 ...
 $ net_od     : num [1:6] 0 0.278 0.522 0.749 0.983 ...
 $ od_per_nmol: num [1:6] NA 0.0348 0.0326 0.0312 0.0307 ...
std_nmol n mean_od sd_od se_od cv_pct median_od net_od od_per_nmol
0 3 0.170 0.0080 0.0046 4.73 0.169 0.000 NA
8 3 0.448 0.0059 0.0034 1.31 0.450 0.278 0.0348
16 3 0.692 0.0060 0.0035 0.87 0.692 0.522 0.0326
24 3 0.918 0.0093 0.0054 1.01 0.921 0.749 0.0312
32 3 1.153 0.0110 0.0063 0.95 1.149 0.983 0.0307
40 3 1.307 0.2082 0.1202 15.93 1.422 1.138 0.0284

GSH standard curve, per-concentration summary at t = 0

Concentrations exceeding 15 % replicate CV: 40 

--- standards_flagged: every standard well with its deviation from the triplicate median ---

tibble [18 × 7] (S3: tbl_df/tbl/data.frame)
 $ well             : chr [1:18] "F5" "F2" "E12" "F4" ...
 $ std_nmol         : num [1:18] 40 32 24 40 8 0 0 16 16 32 ...
 $ od               : num [1:18] 1.067 1.165 0.908 1.433 0.441 ...
 $ source           : chr [1:18] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ triplicate_median: num [1:18] 1.422 1.149 0.921 1.422 0.45 ...
 $ deviation        : num [1:18] -0.355 0.016 -0.013 0.011 -0.009 ...
 $ is_outlier       : logi [1:18] TRUE FALSE FALSE FALSE FALSE FALSE ...
well std_nmol od triplicate_median deviation
F5 40 1.067 1.422 -0.355

Standard wells flagged as outliers (|deviation| > 0.15 A412)

--- fit_comparison: slope/intercept/R^2 for all three candidate fits ---

'data.frame':   3 obs. of  5 variables:
 $ fit      : chr  "all wells" "concentration means" "outlier-excluded"
 $ n        : int  18 6 17
 $ slope    : num  0.0287 0.0287 0.0308
 $ intercept: num  0.208 0.208 0.186
 $ r_squared: num  0.964 0.994 0.999
fit n slope intercept r_squared
all wells 18 0.02868 0.2077 0.96356
concentration means 6 0.02868 0.2077 0.99385
outlier-excluded 17 0.03075 0.1856 0.99852

Standard curve fit comparison (outlier-excluded fit used downstream)

Only one of six concentrations (40 nmol) exceeds the 15% CV threshold — 15.9% across its three replicates — and it is driven entirely by a single well, F5 (A412 1.067 vs. 1.433 and 1.422 for its F4/F6 siblings, a deviation of -0.355 A412 from the triplicate median, well past the 0.15 outlier threshold). Excluding F5, the remaining two 40 nmol replicates agree to 0.5% CV. Every other concentration (0, 8, 16, 24, 32 nmol) is under 5% CV. The outlier-excluded fit reaches R2 = 0.9985 (all-wells fit: R2 = 0.9636).

5.2 Plot the standard curve

std_plot <- ggplot(standards_flagged, aes(x = std_nmol, y = od)) +
  geom_point(aes(colour = is_outlier), size = 3) +
  geom_smooth(data = standards_flagged %>% filter(!is_outlier),
              method = "lm", se = TRUE, colour = "#238b45", fill = "#a1d99b") +
  scale_colour_manual(values = c(`TRUE` = "#cb181d", `FALSE` = "#08519c"),
                      labels = c(`TRUE` = "outlier", `FALSE` = "included"),
                      name = "Standard well") +
  labs(title = "GSH standard curve (t = 0), family F05 36°C",
       subtitle = sprintf("Outlier-excluded fit: slope = %.5f A412/nmol, R² = %.4f (n = %d)",
                          std_slope, std_r2, nrow(standards_flagged %>% filter(!is_outlier))),
       x = "GSH (nmol/well)", y = "A412 at t = 0") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "gsh_standard_curve.png"), std_plot,
       width = 9, height = 6, dpi = 300)

cat("--- std_plot: ggplot object structure ---\n\n")
summary(std_plot)

std_plot

--- std_plot: ggplot object structure ---

data: well, std_nmol, od, source, triplicate_median, deviation,
  is_outlier [18x7]
mapping:  x = ~std_nmol, y = ~od
scales:   colour 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~is_outlier 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

geom_smooth: na.rm = FALSE, orientation = NA, se = TRUE
stat_smooth: na.rm = FALSE, orientation = NA, se = TRUE, method = lm
position_identity

The top two standards (32, 40 nmol) stay within the photometric linear range here — the highest reading (F4, 40 nmol) is A412 1.433, comfortably below the 1.5 ceiling — so the calibration does not need to be restricted to a reduced concentration range.

6 RATE EXTRACTION

Rates are extracted with a sliding-window approach: for each well, the 5-point (8-minute) window of steepest positive slope is taken as the rate, with its R2 retained for the usability check.

well_rates <- compute_well_rates(plate_long, well_diagnostics, assay_params)

cat("--- well_rates: extracted rate window per well ---\n\n")
str(well_rates)
--- well_rates: extracted rate window per well ---

tibble [69 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:69] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:69] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:69] "F05_01_36C" "F05_04_36C" "F05_04_36C" "F05_04_36C" ...
 $ std_nmol              : num [1:69] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:69] 12 12 12 12 12 12 12 12 12 12 ...
 $ t_end                 : num [1:69] 20 20 20 20 20 20 20 20 20 20 ...
 $ slope_mOD_min         : num [1:69] 2.4 2.4 2.5 2.5 2.35 ...
 $ r2                    : num [1:69] 0.998 0.998 1 1 0.999 ...
 $ max_abs_slope_mOD_min : num [1:69] 2.4 2.4 2.5 2.5 2.35 ...
 $ abs_window_is_negative: logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:69] 0.043 0.043 0.044 0.045 0.042 0.041 0.042 0.043 0.042 0.045 ...
 $ frac_rising           : num [1:69] 1 1 1 1 1 1 1 1 1 1 ...
 $ od_max                : num [1:69] 0.222 0.228 0.227 0.228 0.224 0.221 0.226 0.226 0.224 0.227 ...
 $ flag_decreasing       : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:69] 2 8 12 10 14 16 2 10 6 2 ...
 $ flag_bg_active        : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:69] 0 0 0 0 0 0 0 0 0 0 ...
 $ glitch_in_window      : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:69] TRUE TRUE TRUE TRUE TRUE TRUE ...

6.1 Rate tables by well type

sample_rates <- well_rates %>% filter(well_type == "sample")

kable(sample_rates %>% select(well, sample_id, t_start, t_end, slope_mOD_min, r2, rate_usable),
      digits = c(0, 0, 0, 0, 3, 4, 0),
      caption = "Extracted rate window, sample wells")

cat("\nSample wells with a usable rate:", sum(sample_rates$rate_usable), "/", nrow(sample_rates), "\n")
cat("Sample rate range (mA412/min):",
    paste(sprintf("%.2f", range(sample_rates$slope_mOD_min)), collapse = " - "), "\n")
cat("Median replicate-level R^2:", sprintf("%.4f", median(sample_rates$r2)), "\n")

cat("\n--- sample_rates: rate-window results restricted to sample wells ---\n\n")
str(sample_rates)
well sample_id t_start t_end slope_mOD_min r2 rate_usable
A1 F05_01_36C 12 20 2.40 0.9983 TRUE
A10 F05_04_36C 12 20 2.40 0.9983 TRUE
A11 F05_04_36C 12 20 2.50 1.0000 TRUE
A12 F05_04_36C 12 20 2.50 1.0000 TRUE
A2 F05_01_36C 12 20 2.35 0.9986 TRUE
A3 F05_01_36C 12 20 2.25 0.9985 TRUE
A4 F05_02_36C 12 20 2.40 0.9983 TRUE
A5 F05_02_36C 12 20 2.40 0.9983 TRUE
A6 F05_02_36C 12 20 2.30 0.9981 TRUE
A7 F05_03_36C 12 20 2.50 1.0000 TRUE
A8 F05_03_36C 12 20 2.35 0.9986 TRUE
A9 F05_03_36C 12 20 2.50 1.0000 TRUE
B1 F05_05_36C 12 20 2.35 0.9986 TRUE
B10 F05_08_36C 12 20 2.30 0.9944 TRUE
B11 F05_08_36C 12 20 2.35 0.9986 TRUE
B12 F05_08_36C 12 20 2.65 0.9989 TRUE
B2 F05_05_36C 12 20 2.30 0.9981 TRUE
B3 F05_05_36C 12 20 2.35 0.9986 TRUE
B4 F05_06_36C 12 20 2.40 0.9983 TRUE
B5 F05_06_36C 12 20 2.40 0.9983 TRUE
B6 F05_06_36C 12 20 2.30 0.9981 TRUE
B7 F05_07_36C 12 20 2.35 0.9986 TRUE
B8 F05_07_36C 12 20 2.40 0.9983 TRUE
B9 F05_07_36C 12 20 2.35 0.9986 TRUE

Extracted rate window, sample wells

Sample wells with a usable rate: 24 / 24 
Sample rate range (mA412/min): 2.25 - 2.65 
Median replicate-level R^2: 0.9986 

--- sample_rates: rate-window results restricted to sample wells ---

tibble [24 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:24] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:24] "F05_01_36C" "F05_04_36C" "F05_04_36C" "F05_04_36C" ...
 $ std_nmol              : num [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:24] 12 12 12 12 12 12 12 12 12 12 ...
 $ t_end                 : num [1:24] 20 20 20 20 20 20 20 20 20 20 ...
 $ slope_mOD_min         : num [1:24] 2.4 2.4 2.5 2.5 2.35 ...
 $ r2                    : num [1:24] 0.998 0.998 1 1 0.999 ...
 $ max_abs_slope_mOD_min : num [1:24] 2.4 2.4 2.5 2.5 2.35 ...
 $ abs_window_is_negative: logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:24] 0.043 0.043 0.044 0.045 0.042 0.041 0.042 0.043 0.042 0.045 ...
 $ frac_rising           : num [1:24] 1 1 1 1 1 1 1 1 1 1 ...
 $ od_max                : num [1:24] 0.222 0.228 0.227 0.228 0.224 0.221 0.226 0.226 0.224 0.227 ...
 $ flag_decreasing       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:24] 2 8 12 10 14 16 2 10 6 2 ...
 $ flag_bg_active        : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:24] 0 0 0 0 0 0 0 0 0 0 ...
 $ glitch_in_window      : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:24] TRUE TRUE TRUE TRUE TRUE TRUE ...

Every one of the 24 sample wells clears the R2 >= 0.80 and positive-net-change usability bar; in fact every replicate-level R2 exceeds 0.99. No well needed to be excluded from rate extraction on this plate — a direct consequence of the clean, monotonically-rising traces noted above.

6.2 Positive control

pos_control <- compute_positive_control(well_rates, std_curve$std_slope, assay_params)

kable(pos_control %>% select(well, t_start, t_end, slope_mOD_min, r2, net_change, activity_mU_uL),
      digits = c(0, 0, 0, 3, 4, 3, 4),
      caption = "Positive control replicates")

cat("Positive control mean rate:", sprintf("%.3f", mean(pos_control$slope_mOD_min)), "mA412/min\n")
cat("Positive control CV:", sprintf("%.1f",
    100 * sd(pos_control$slope_mOD_min) / mean(pos_control$slope_mOD_min)), "%\n")
cat("All replicates rising and linear (R^2 > 0.99):",
    all(pos_control$net_change > 0 & pos_control$r2 > 0.99), "\n")

cat("\n--- pos_control: positive control rate and R^2 per replicate ---\n\n")
str(pos_control)
well t_start t_end slope_mOD_min r2 net_change activity_mU_uL
F7 12 20 2.8 0.9987 0.050 0.0455
F8 12 20 3.0 1.0000 0.053 0.0488
F9 12 20 3.0 1.0000 0.054 0.0488

Positive control replicates

Positive control mean rate: 2.933 mA412/min
Positive control CV: 3.9 %
All replicates rising and linear (R^2 > 0.99): TRUE 

--- pos_control: positive control rate and R^2 per replicate ---

tibble [3 × 22] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:3] "F7" "F8" "F9"
 $ well_type             : chr [1:3] "positive_control" "positive_control" "positive_control"
 $ sample_id             : chr [1:3] NA NA NA
 $ std_nmol              : num [1:3] NA NA NA
 $ t_start               : num [1:3] 12 12 12
 $ t_end                 : num [1:3] 20 20 20
 $ slope_mOD_min         : num [1:3] 2.8 3 3
 $ r2                    : num [1:3] 0.999 1 1
 $ max_abs_slope_mOD_min : num [1:3] 2.8 3 3
 $ abs_window_is_negative: logi [1:3] FALSE FALSE FALSE
 $ net_change            : num [1:3] 0.05 0.053 0.054
 $ frac_rising           : num [1:3] 1 1 1
 $ od_max                : num [1:3] 0.301 0.303 0.306
 $ flag_decreasing       : logi [1:3] FALSE FALSE FALSE
 $ flag_over_range       : logi [1:3] FALSE FALSE FALSE
 $ flag_glitch           : logi [1:3] FALSE FALSE FALSE
 $ glitch_at_min         : num [1:3] 2 2 12
 $ flag_bg_active        : logi [1:3] FALSE FALSE FALSE
 $ n_flags               : int [1:3] 0 0 0
 $ glitch_in_window      : logi [1:3] FALSE FALSE FALSE
 $ rate_usable           : logi [1:3] TRUE TRUE TRUE
 $ activity_mU_uL        : num [1:3] 0.0455 0.0488 0.0488

The positive control again performed exactly as expected: all three replicates rise linearly (R2 > 0.998) at 2.8-3.0 mA412/min with low replicate CV, confirming the core reaction chemistry and reader were functioning normally on this plate.

7 BACKGROUND CORRECTION

The background rate is estimated from only the flat (well-behaved) background replicates, since a well flagged flag_bg_active does not measure background.

background_per_sample <- compute_background_correction(well_rates)

cat("Background rate from flat wells only (mA412/min):\n")
print(background_per_sample %>% as.data.frame(), row.names = FALSE, digits = 3)

cat("\nMedian flat-background rate:",
    sprintf("%.3f", median(background_per_sample$bg_rate_flat, na.rm = TRUE)),
    "mA412/min\n")
cat("Median sample rate:",
    sprintf("%.2f", median(sample_rates$slope_mOD_min)), "mA412/min\n")
cat("Background as % of sample signal:",
    sprintf("%.1f", 100 * median(background_per_sample$bg_rate_flat, na.rm = TRUE) /
                    median(sample_rates$slope_mOD_min)), "%\n")

cat("\n--- background_per_sample: per-sample background rate estimates ---\n\n")
str(background_per_sample)
Background rate from flat wells only (mA412/min):
  sample_id n_bg_total n_bg_flat bg_rate_flat bg_rate_all
 F05_01_36C          3         3        0.167       0.167
 F05_02_36C          3         3        0.150       0.150
 F05_03_36C          3         2        0.150      -0.433
 F05_04_36C          3         3        0.183       0.183
 F05_05_36C          3         3        0.183       0.183
 F05_06_36C          3         3        0.233       0.233
 F05_07_36C          3         3        0.167       0.167
 F05_08_36C          3         3        0.183       0.183

Median flat-background rate: 0.175 mA412/min
Median sample rate: 2.37 mA412/min
Background as % of sample signal: 7.4 %

--- background_per_sample: per-sample background rate estimates ---

tibble [8 × 5] (S3: tbl_df/tbl/data.frame)
 $ sample_id   : chr [1:8] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ n_bg_total  : int [1:8] 3 3 3 3 3 3 3 3
 $ n_bg_flat   : int [1:8] 3 3 2 3 3 3 3 3
 $ bg_rate_flat: num [1:8] 0.167 0.15 0.15 0.183 0.183 ...
 $ bg_rate_all : num [1:8] 0.167 0.15 -0.433 0.183 0.183 ...

Background correction remains a minor term (~7% of sample signal), and the one active-background well (C8, F05_03_36C) simply drops from that sample’s flat-well average without leaving any sample without a background estimate.

7.1 Background significance test

The flat/active threshold classification above is backed by an explicit significance test: a Welch two-sample t-test of well-level rate (background vs. sample), plus an ANOVA on the OD trajectory over time. This mirrors the one-time background ANOVA validation done by Cattau et al. (2023) on the legacy version of this assay.

bg_sample_rates <- well_rates %>%
  filter(well_type %in% c("sample", "background"))

bg_significance_test <- t.test(slope_mOD_min ~ well_type, data = bg_sample_rates)

cat("--- Welch two-sample t-test: well-level rate (background vs. sample) ---\n\n")
print(bg_significance_test)

bg_trend_anova <- anova(lm(od ~ well_type * time_min,
                            data = plate_long %>% filter(well_type %in% c("sample", "background"))))

cat("\n--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---\n\n")
print(bg_trend_anova)

cat("\nInterpretation: mean background-well rate is",
    sprintf("%.2f", unname(bg_significance_test$estimate["mean in group background"])),
    "mA412/min vs.", sprintf("%.2f", unname(bg_significance_test$estimate["mean in group sample"])),
    "mA412/min for sample wells (p =", format.pval(bg_significance_test$p.value, digits = 3),
    "). Background wells are statistically distinct from, and far slower\n",
    "than, reaction wells on this plate -- consistent with the per-well\n",
    "flat/active classification above.\n")

cat("\n--- bg_significance_test: Welch t-test result object ---\n\n")
str(bg_significance_test)

cat("\n--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---\n\n")
str(bg_trend_anova)
--- Welch two-sample t-test: well-level rate (background vs. sample) ---


    Welch Two Sample t-test

data:  slope_mOD_min by well_type
t = -29.833, df = 25.695, p-value < 2.2e-16
alternative hypothesis: true difference in means between group background and group sample is not equal to 0
95 percent confidence interval:
 -2.442975 -2.127858
sample estimates:
mean in group background     mean in group sample 
               0.1041667                2.3895833 


--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---

Analysis of Variance Table

Response: od
                    Df  Sum Sq Mean Sq  F value    Pr(>F)    
well_type            1 2.50292 2.50292 4026.991 < 2.2e-16 ***
time_min             1 0.02404 0.02404   38.683 1.020e-09 ***
well_type:time_min   1 0.02396 0.02396   38.545 1.089e-09 ***
Residuals          524 0.32568 0.00062                       
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Interpretation: mean background-well rate is 0.10 mA412/min vs. 2.39 mA412/min for sample wells (p = <2e-16 ). Background wells are statistically distinct from, and far slower
 than, reaction wells on this plate -- consistent with the per-well
 flat/active classification above.

--- bg_significance_test: Welch t-test result object ---

List of 10
 $ statistic  : Named num -29.8
  ..- attr(*, "names")= chr "t"
 $ parameter  : Named num 25.7
  ..- attr(*, "names")= chr "df"
 $ p.value    : num 1.82e-21
 $ conf.int   : num [1:2] -2.44 -2.13
  ..- attr(*, "conf.level")= num 0.95
 $ estimate   : Named num [1:2] 0.104 2.39
  ..- attr(*, "names")= chr [1:2] "mean in group background" "mean in group sample"
 $ null.value : Named num 0
  ..- attr(*, "names")= chr "difference in means between group background and group sample"
 $ stderr     : num 0.0766
 $ alternative: chr "two.sided"
 $ method     : chr "Welch Two Sample t-test"
 $ data.name  : chr "slope_mOD_min by well_type"
 - attr(*, "class")= chr "htest"

--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---

Classes 'anova' and 'data.frame':   4 obs. of  5 variables:
 $ Df     : int  1 1 1 524
 $ Sum Sq : num  2.503 0.024 0.024 0.326
 $ Mean Sq: num  2.502918 0.024043 0.023957 0.000622
 $ F value: num  4027 38.7 38.5 NA
 $ Pr(>F) : num  4.07e-248 1.02e-09 1.09e-09 NA
 - attr(*, "heading")= chr [1:2] "Analysis of Variance Table\n" "Response: od"

8 TECHNICAL REPLICATE PRECISION

Coefficient of variation across the three technical replicates of each sample, computed on the extracted rate. Both all and usable replicates only are reported; here they are identical for every sample because no replicate was excluded.

cv_summary <- compute_replicate_cv(well_rates, protein_by_sample$sample_id, assay_params)

kable(cv_summary %>% select(sample_id, n_all, mean_all, sd_all, cv_all,
                            n_usable, mean_usable, sd_usable, cv_usable,
                            excluded_wells),
      digits = c(0, 0, 2, 2, 1, 0, 2, 2, 1, 0),
      col.names = c("Sample", "n", "Mean rate", "SD", "CV (%)",
                    "n usable", "Mean rate", "SD", "CV (%)", "Excluded wells"),
      caption = paste0("Technical replicate CV of the extracted rate (mA412/min), ",
                       "all replicates vs. usable replicates only"))

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on ALL replicates:", sum(cv_summary$fails_cv_all), "/", nrow(cv_summary), "\n")
print(cv_summary %>% filter(fails_cv_all) %>%
        select(sample_id, n_all, mean_all, sd_all, cv_all) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on USABLE replicates:", sum(cv_summary$fails_cv_usable), "\n")
print(cv_summary %>% filter(fails_cv_usable) %>%
        select(sample_id, n_usable, mean_usable, sd_usable, cv_usable) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

write.csv(cv_summary, file.path(output_dir, "technical_replicate_cv.csv"), row.names = FALSE)

cat("\n--- cv_summary: per-sample technical replicate CV, both variants ---\n\n")
str(cv_summary)
Sample n Mean rate SD CV (%) n usable Mean rate SD CV (%) Excluded wells
F05_01_36C 3 2.33 0.08 3.3 3 2.33 0.08 3.3 -
F05_02_36C 3 2.37 0.06 2.4 3 2.37 0.06 2.4 -
F05_03_36C 3 2.45 0.09 3.5 3 2.45 0.09 3.5 -
F05_04_36C 3 2.47 0.06 2.3 3 2.47 0.06 2.3 -
F05_05_36C 3 2.33 0.03 1.2 3 2.33 0.03 1.2 -
F05_06_36C 3 2.37 0.06 2.4 3 2.37 0.06 2.4 -
F05_07_36C 3 2.37 0.03 1.2 3 2.37 0.03 1.2 -
F05_08_36C 3 2.43 0.19 7.8 3 2.43 0.19 7.8 -

Technical replicate CV of the extracted rate (mA412/min), all replicates vs. usable replicates only

Samples exceeding CV 15% on ALL replicates: 0 / 8 
[1] sample_id n_all     mean_all  sd_all    cv_all   
<0 rows> (or 0-length row.names)

Samples exceeding CV 15% on USABLE replicates: 0 
[1] sample_id   n_usable    mean_usable sd_usable   cv_usable  
<0 rows> (or 0-length row.names)

--- cv_summary: per-sample technical replicate CV, both variants ---

'data.frame':   8 obs. of  12 variables:
 $ sample_id      : chr  "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ n_all          : int  3 3 3 3 3 3 3 3
 $ mean_all       : num  2.33 2.37 2.45 2.47 2.33 ...
 $ sd_all         : num  0.0764 0.0577 0.0866 0.0577 0.0289 ...
 $ cv_all         : num  3.27 2.44 3.53 2.34 1.24 ...
 $ n_usable       : int  3 3 3 3 3 3 3 3
 $ mean_usable    : num  2.33 2.37 2.45 2.47 2.33 ...
 $ sd_usable      : num  0.0764 0.0577 0.0866 0.0577 0.0289 ...
 $ cv_usable      : num  3.27 2.44 3.53 2.34 1.24 ...
 $ excluded_wells : Named chr  "-" "-" "-" "-" ...
  ..- attr(*, "names")= chr [1:8] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ fails_cv_all   : logi  FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ fails_cv_usable: logi  FALSE FALSE FALSE FALSE FALSE FALSE ...

No sample exceeds the 15% CV threshold on this plate. The highest is F05_08_36C at 7.8%; every other sample is under 4%.

8.1 Plot replicate spread

cv_plot_data <- sample_rates %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id")

cv_plot <- ggplot(cv_plot_data, aes(x = sample_id, y = slope_mOD_min)) +
  geom_point(aes(colour = rate_usable, shape = rate_usable),
             size = 3, position = position_jitter(width = 0.12, height = 0)) +
  stat_summary(fun = mean, geom = "crossbar", width = 0.5,
               colour = "grey30", linewidth = 0.3) +
  geom_text(data = cv_summary, aes(x = sample_id, y = Inf,
                                   label = sprintf("CV %.0f%%", cv_all)),
            vjust = 1.6, size = 3,
            colour = ifelse(cv_summary$fails_cv_all, "#cb181d", "grey30")) +
  scale_colour_manual(values = c(`TRUE` = "#08519c", `FALSE` = "#cb181d"),
                      labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                     labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  labs(title = "Technical replicate spread in extracted rate",
       subtitle = paste0("Crossbar = mean of all three replicates; CV computed on all three. ",
                         "Threshold ", assay_params$cv_threshold_pct, "%."),
       x = "Sample", y = "Rate (mA412/min)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 30, hjust = 1))

ggsave(file.path(output_dir, "technical_replicate_cv.png"), cv_plot,
       width = 9, height = 6, dpi = 300)

cat("--- cv_plot: ggplot object structure ---\n\n")
summary(cv_plot)

cv_plot

--- cv_plot: ggplot object structure ---

data: well, well_type, sample_id, std_nmol, t_start, t_end,
  slope_mOD_min, r2, max_abs_slope_mOD_min, abs_window_is_negative,
  net_change, frac_rising, od_max, flag_decreasing, flag_over_range,
  flag_glitch, glitch_at_min, flag_bg_active, n_flags,
  glitch_in_window, rate_usable, cv_all [24x22]
mapping:  x = ~sample_id, y = ~slope_mOD_min
scales:   colour, shape 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~rate_usable, shape = ~rate_usable 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_jitter 

geom_crossbar: na.rm = FALSE, orientation = NA, width = 0.5
stat_summary: fun.data = NULL, fun = function (x, ...) 
UseMethod("mean"), fun.max = NULL, fun.min = NULL, fun.args = list(), na.rm = FALSE, orientation = NA
position_identity 

mapping: x = ~sample_id, y = Inf, label = ~sprintf("CV %.0f%%", cv_all) 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge

9 CITRATE SYNTHASE ACTIVITY

9.1 Calculation

Following Abcam Section 10.3:

Sample CS activity = B / (dT x V) x D nmol/min/µL (= mU/µL)

  1. Rate (A412/min) = max-increasing sliding-window slope / 1000
  2. Background-corrected rate = sample rate - mean flat background rate
  3. B/dT (nmol/min) = background-corrected rate / standard curve slope (0.03075 A412 per nmol, from the outlier-excluded fit)
  4. Activity (mU/µL) = (B/dT) / V x D, with V = 2 µL and D = 1
  5. Protein-normalized (mU/mg protein) = activity x homogenate_volume_uL (350 µL, measured) / total protein (mg)
cs_activity <- calculate_cs_activity(well_rates, protein_by_sample, background_per_sample,
                                      std_slope, std_nmol_max, assay_params, plate_long)

cat("Samples with a computed activity:", nrow(cs_activity), "/", nrow(protein_by_sample), "\n")
cat("Activity range (mU/uL):",
    paste(sprintf("%.4f", range(cs_activity$activity_mU_per_uL)), collapse = " - "), "\n")
cat("Activity range (mU/mg protein):",
    paste(sprintf("%.3f", range(cs_activity$activity_mU_per_mg_protein)), collapse = " - "), "\n")

cat("\n--- cs_activity: per-sample citrate synthase activity ---\n\n")
str(cs_activity)
Samples with a computed activity: 8 / 8 
Activity range (mU/uL): 0.0347 - 0.0374 
Activity range (mU/mg protein): 20.972 - 66.031 

--- cs_activity: per-sample citrate synthase activity ---

tibble [8 × 21] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:8] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ n_reps_used               : int [1:8] 3 3 3 3 3 3 3 3
 $ mean_rate_mOD_min         : num [1:8] 2.33 2.37 2.45 2.47 2.33 ...
 $ sd_rate                   : num [1:8] 0.0764 0.0577 0.0866 0.0577 0.0289 ...
 $ cv_rate                   : num [1:8] 3.27 2.44 3.53 2.34 1.24 ...
 $ family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ individual                : chr [1:8] "01" "02" "03" "04" ...
 $ temperature               : chr [1:8] "36C" "36C" "36C" "36C" ...
 $ conc_ug_mL                : num [1:8] 953 546 668 1062 1667 ...
 $ total_protein_mg          : num [1:8] 0.334 0.191 0.234 0.372 0.583 ...
 $ bg_rate_flat              : num [1:8] 0.167 0.15 0.15 0.183 0.183 ...
 $ n_bg_flat                 : int [1:8] 3 3 2 3 3 3 3 3
 $ bg_rate_mOD_min           : num [1:8] 0.167 0.15 0.15 0.183 0.183 ...
 $ corrected_mOD_min         : num [1:8] 2.17 2.22 2.3 2.28 2.15 ...
 $ rate_OD_min               : num [1:8] 0.00217 0.00222 0.0023 0.00228 0.00215 ...
 $ nmol_per_min              : num [1:8] 0.0705 0.0721 0.0748 0.0742 0.0699 ...
 $ activity_mU_per_uL        : num [1:8] 0.0352 0.036 0.0374 0.0371 0.035 ...
 $ total_mU_in_homogenate    : num [1:8] 12.3 12.6 13.1 13 12.2 ...
 $ activity_mU_per_mg_protein: num [1:8] 37 66 56 34.9 21 ...
 $ nmol_in_window            : num [1:8] 1.41 1.44 1.5 1.48 1.4 ...
 $ within_std_range          : logi [1:8] TRUE TRUE TRUE TRUE TRUE TRUE ...

9.2 Results table

results_table <- build_results_table(cs_activity, cv_summary, baseline_per_sample, assay_params)

kable(results_table,
      caption = paste0("Citrate synthase activity, family F05 36°C (mU/mg protein uses ",
                       "measured protein concentration x the ", assay_params$homogenate_volume_uL,
                       " uL homogenization volume)"))

cat("\nInterpretation summary:\n")
print(results_table %>% count(Interpretation, name = "n_samples") %>%
        as.data.frame(), row.names = FALSE)

write.csv(results_table, file.path(output_dir, "citrate_synthase_activity_results.csv"),
          row.names = FALSE)

cat("\n--- results_table: formatted per-sample results written to CSV ---\n\n")
str(results_table)
Sample Family Individual Temperature Protein conc (ug/mL) Total protein (mg) Reps used CV all reps (%) CV used reps (%) Rate (mA412/min) BG rate (mA412/min) Corrected rate (mA412/min) Activity (mU/uL) Activity (mU/mg protein) CV flag Elevated baseline reps Interpretation
F05_01_36C F05 01 36C 953.1 0.334 3/3 3.3 3.3 2.33 0.167 2.17 0.0352 36.960 pass 0/3 usable
F05_02_36C F05 02 36C 545.8 0.191 3/3 2.4 2.4 2.37 0.150 2.22 0.0360 66.031 pass 0/3 usable
F05_03_36C F05 03 36C 667.6 0.234 3/3 3.5 3.5 2.45 0.150 2.30 0.0374 56.014 pass 0/3 usable
F05_04_36C F05 04 36C 1062.2 0.372 3/3 2.3 2.3 2.47 0.183 2.28 0.0371 34.950 pass 0/3 usable
F05_05_36C F05 05 36C 1666.8 0.583 3/3 1.2 1.2 2.33 0.183 2.15 0.0350 20.972 pass 0/3 usable
F05_06_36C F05 06 36C 1265.9 0.443 3/3 2.4 2.4 2.37 0.233 2.13 0.0347 27.399 pass 0/3 usable
F05_07_36C F05 07 36C 1110.5 0.389 3/3 1.2 1.2 2.37 0.167 2.20 0.0358 32.210 pass 0/3 usable
F05_08_36C F05 08 36C 1357.5 0.475 3/3 7.8 7.8 2.43 0.183 2.25 0.0366 26.948 pass 0/3 usable

Citrate synthase activity, family F05 36°C (mU/mg protein uses measured protein concentration x the 350 uL homogenization volume)

Interpretation summary:
 Interpretation n_samples
         usable         8

--- results_table: formatted per-sample results written to CSV ---

tibble [8 × 17] (S3: tbl_df/tbl/data.frame)
 $ Sample                    : chr [1:8] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ Family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ Individual                : chr [1:8] "01" "02" "03" "04" ...
 $ Temperature               : chr [1:8] "36C" "36C" "36C" "36C" ...
 $ Protein conc (ug/mL)      : num [1:8] 953 546 668 1062 1667 ...
 $ Total protein (mg)        : num [1:8] 0.334 0.191 0.234 0.372 0.583 0.443 0.389 0.475
 $ Reps used                 : chr [1:8] "3/3" "3/3" "3/3" "3/3" ...
 $ CV all reps (%)           : num [1:8] 3.3 2.4 3.5 2.3 1.2 2.4 1.2 7.8
 $ CV used reps (%)          : num [1:8] 3.3 2.4 3.5 2.3 1.2 2.4 1.2 7.8
 $ Rate (mA412/min)          : num [1:8] 2.33 2.37 2.45 2.47 2.33 2.37 2.37 2.43
 $ BG rate (mA412/min)       : num [1:8] 0.167 0.15 0.15 0.183 0.183 0.233 0.167 0.183
 $ Corrected rate (mA412/min): num [1:8] 2.17 2.22 2.3 2.28 2.15 2.13 2.2 2.25
 $ Activity (mU/uL)          : num [1:8] 0.0352 0.036 0.0374 0.0371 0.035 0.0347 0.0358 0.0366
 $ Activity (mU/mg protein)  : num [1:8] 37 66 56 35 21 ...
 $ CV flag                   : chr [1:8] "pass" "pass" "pass" "pass" ...
 $ Elevated baseline reps    : chr [1:8] "0/3" "0/3" "0/3" "0/3" ...
 $ Interpretation            : chr [1:8] "usable" "usable" "usable" "usable" ...

All eight samples are usable on this plate — no CV flag, no elevated baseline replicate.

9.3 Plot activity

# Three-level quality scheme, kept for consistency with other plate
# documents even though no sample here falls in the flagged categories.
qual_levels <- c("clean", paste0("CV > ", assay_params$cv_threshold_pct, "%"),
                 "all reps baseline-compromised")
qual_cols   <- setNames(c("#08519c", "#cb181d", "grey60"), qual_levels)

activity_plot_data <- cs_activity %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id") %>%
  left_join(baseline_per_sample %>% select(sample_id, n_elevated, n), by = "sample_id") %>%
  mutate(cv_flag = factor(case_when(
                            n_elevated == n                              ~ qual_levels[3],
                            cv_all > assay_params$cv_threshold_pct       ~ qual_levels[2],
                            TRUE                                        ~ qual_levels[1]),
                          levels = qual_levels),
         label_n = paste0(n_reps_used, "/3 reps",
                          ifelse(n_elevated == n, " - DISQUALIFIED", "")))

activity_plot <- ggplot(activity_plot_data,
                        aes(x = reorder(sample_id, activity_mU_per_uL),
                            y = activity_mU_per_uL, fill = cv_flag)) +
  geom_col(width = 0.65) +
  geom_text(aes(label = label_n), hjust = -0.15, size = 3, colour = "grey25") +
  scale_fill_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  coord_flip(clip = "off") +
  expand_limits(y = max(activity_plot_data$activity_mU_per_uL) * 1.25) +
  labs(title = "Citrate synthase activity, M. gigas ctenidia, family F05 36°C",
       subtitle = "Background-corrected, mean of all three technical replicates (all usable on this plate).",
       x = NULL, y = "CS activity (mU/uL homogenate)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "citrate_synthase_activity.png"), activity_plot,
       width = 10, height = 6, dpi = 300)

cat("--- activity_plot: ggplot object structure ---\n\n")
summary(activity_plot)

activity_plot

--- activity_plot: ggplot object structure ---

data: sample_id, n_reps_used, mean_rate_mOD_min, sd_rate, cv_rate,
  family, individual, temperature, conc_ug_mL, total_protein_mg,
  bg_rate_flat, n_bg_flat, bg_rate_mOD_min, corrected_mOD_min,
  rate_OD_min, nmol_per_min, activity_mU_per_uL,
  total_mU_in_homogenate, activity_mU_per_mg_protein, nmol_in_window,
  within_std_range, cv_all, n_elevated, n, cv_flag, label_n [8x26]
mapping:  x = ~reorder(sample_id, activity_mU_per_uL), y = ~activity_mU_per_uL, fill = ~cv_flag
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: label = ~label_n 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge 

mapping: y = ~y 
geom_blank: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

Raw activity (mU/µL) is remarkably uniform across all eight individuals — 0.0347 to 0.0374, a spread of only 7.8%, against a technical CV of under 8% for any single sample.

9.4 Protein-normalized activity

# All eight samples are clean on this plate, so the trend fit uses all of them.
norm_clean <- activity_plot_data %>% filter(cv_flag == qual_levels[1])

norm_plot <- ggplot(activity_plot_data,
                    aes(x = total_protein_mg, y = activity_mU_per_mg_protein)) +
  geom_smooth(data = norm_clean, method = "lm", se = TRUE,
              colour = "grey35", fill = "grey85", linewidth = 0.6) +
  geom_point(aes(colour = cv_flag), size = 3.5) +
  geom_text(aes(label = individual), vjust = -1, size = 3, colour = "grey25") +
  scale_colour_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  labs(title = "Protein-normalized CS activity vs. total extracted protein",
       subtitle = paste0("Point labels = individual ID. Total protein = measured protein ",
                         "concentration x the same\n", assay_params$homogenate_volume_uL,
                         " uL homogenization volume used for every sample. Grey line = fit to ",
                         "all 8\nsamples (all clean on this plate); a negative slope would indicate the ",
                         "normalization\nis not protein-independent."),
       x = "Total extracted protein (mg)", y = "CS activity (mU/mg protein)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

protein_trend <- cor.test(activity_plot_data$total_protein_mg, activity_plot_data$activity_mU_per_mg_protein)
protein_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_mg_protein)
uL_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_uL)

ggsave(file.path(output_dir, "citrate_synthase_activity_normalized.png"), norm_plot,
       width = 10, height = 6, dpi = 300)

cat("Correlation of mU/mg protein with total extracted protein\n")
cat(sprintf("  all %d samples      : r = %+.3f, p = %.4f\n",
            nrow(activity_plot_data), protein_trend$estimate, protein_trend$p.value))
cat(sprintf("  clean samples only  : r = %+.3f, p = %.4f  (n = %d)\n",
            protein_trend_clean$estimate, protein_trend_clean$p.value, nrow(norm_clean)))
cat(sprintf("  clean-sample mU/mg protein range: %.3f to %.3f (%.1f-fold)\n",
            min(norm_clean$activity_mU_per_mg_protein), max(norm_clean$activity_mU_per_mg_protein),
            max(norm_clean$activity_mU_per_mg_protein) / min(norm_clean$activity_mU_per_mg_protein)))
cat(sprintf("  clean-sample mU/uL range: %.4f to %.4f (%.1f-fold)\n",
            min(norm_clean$activity_mU_per_uL), max(norm_clean$activity_mU_per_uL),
            max(norm_clean$activity_mU_per_uL) / min(norm_clean$activity_mU_per_uL)))
cat(sprintf("\nArtifact check: r(total protein, mU/uL raw rate) = %+.3f, p = %.4f  (n = %d)\n",
            uL_trend_clean$estimate, uL_trend_clean$p.value, nrow(norm_clean)))
cat("If mU/uL itself trends with total protein while mU/mg protein trends oppositely,\n",
    "the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate\n",
    "rather than a biological difference in specific activity.\n", sep = "")

cat("\n--- protein_trend_clean: correlation test on clean samples ---\n\n")
str(protein_trend_clean, max.level = 1, give.attr = FALSE)

cat("\n--- norm_clean: clean samples used for the trend fit ---\n\n")
str(norm_clean)

norm_plot

Correlation of mU/mg protein with total extracted protein
  all 8 samples      : r = -0.948, p = 0.0003
  clean samples only  : r = -0.948, p = 0.0003  (n = 8)
  clean-sample mU/mg protein range: 20.972 to 66.031 (3.1-fold)
  clean-sample mU/uL range: 0.0347 to 0.0374 (1.1-fold)

Artifact check: r(total protein, mU/uL raw rate) = -0.479, p = 0.2302  (n = 8)
If mU/uL itself trends with total protein while mU/mg protein trends oppositely,
the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate
rather than a biological difference in specific activity.

--- protein_trend_clean: correlation test on clean samples ---

List of 9
 $ statistic  : Named num -7.31
 $ parameter  : Named int 6
 $ p.value    : num 0.000335
 $ estimate   : Named num -0.948
 $ null.value : Named num 0
 $ alternative: chr "two.sided"
 $ method     : chr "Pearson's product-moment correlation"
 $ data.name  : chr "norm_clean$total_protein_mg and norm_clean$activity_mU_per_mg_protein"
 $ conf.int   : num [1:2] -0.991 -0.734

--- norm_clean: clean samples used for the trend fit ---

tibble [8 × 26] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:8] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ n_reps_used               : int [1:8] 3 3 3 3 3 3 3 3
 $ mean_rate_mOD_min         : num [1:8] 2.33 2.37 2.45 2.47 2.33 ...
 $ sd_rate                   : num [1:8] 0.0764 0.0577 0.0866 0.0577 0.0289 ...
 $ cv_rate                   : num [1:8] 3.27 2.44 3.53 2.34 1.24 ...
 $ family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ individual                : chr [1:8] "01" "02" "03" "04" ...
 $ temperature               : chr [1:8] "36C" "36C" "36C" "36C" ...
 $ conc_ug_mL                : num [1:8] 953 546 668 1062 1667 ...
 $ total_protein_mg          : num [1:8] 0.334 0.191 0.234 0.372 0.583 ...
 $ bg_rate_flat              : num [1:8] 0.167 0.15 0.15 0.183 0.183 ...
 $ n_bg_flat                 : int [1:8] 3 3 2 3 3 3 3 3
 $ bg_rate_mOD_min           : num [1:8] 0.167 0.15 0.15 0.183 0.183 ...
 $ corrected_mOD_min         : num [1:8] 2.17 2.22 2.3 2.28 2.15 ...
 $ rate_OD_min               : num [1:8] 0.00217 0.00222 0.0023 0.00228 0.00215 ...
 $ nmol_per_min              : num [1:8] 0.0705 0.0721 0.0748 0.0742 0.0699 ...
 $ activity_mU_per_uL        : num [1:8] 0.0352 0.036 0.0374 0.0371 0.035 ...
 $ total_mU_in_homogenate    : num [1:8] 12.3 12.6 13.1 13 12.2 ...
 $ activity_mU_per_mg_protein: num [1:8] 37 66 56 34.9 21 ...
 $ nmol_in_window            : num [1:8] 1.41 1.44 1.5 1.48 1.4 ...
 $ within_std_range          : logi [1:8] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ cv_all                    : num [1:8] 3.27 2.44 3.53 2.34 1.24 ...
 $ n_elevated                : int [1:8] 0 0 0 0 0 0 0 0
 $ n                         : int [1:8] 3 3 3 3 3 3 3 3
 $ cv_flag                   : Factor w/ 3 levels "clean","CV > 15%",..: 1 1 1 1 1 1 1 1
 $ label_n                   : chr [1:8] "3/3 reps" "3/3 reps" "3/3 reps" "3/3 reps" ...

mU/mg protein correlates negatively with total extracted protein at r = -0.948 (p = 0.0003, n = 8, all 8 samples clean on this plate), while raw mU/µL is essentially flat against protein (r = -0.479, p = 0.23). Because the raw signal used in the numerator is close to protein-independent here, the negative mU/mg protein trend on this plate is substantially closer to a pure division artifact (dividing a near-constant numerator by a 3-fold-varying denominator) than a genuine biological effect. This distinction matters directly for note 3 of the recommendations below.

10 QC SUMMARY

qc_summary <- build_qc_summary_table(layout_wells, recon$summary, std_curve,
                                      well_diagnostics, well_rates, cv_summary,
                                      baseline_per_sample, pos_control, assay_params)

kable(qc_summary, col.names = c("QC check", "Result"),
      caption = "Quality control summary for this plate")

write.csv(qc_summary, file.path(output_dir, "qc_summary.csv"), row.names = FALSE)

cat("\n--- qc_summary: consolidated QC checks for the run ---\n\n")
str(qc_summary)
QC check Result
Occupied wells in layout 69
Layout wells missing from absorbance CSV 0
Max disagreement between absorbance CSV and full report 0
Standard concentrations with replicate CV > threshold 40
Standard wells flagged as outliers 1/18 (F5)
Standard curve R^2 (outlier-excluded, replicate level) 0.9985
Standard curve R^2 (all wells, replicate level) 0.9636
Standard wells that LOST signal over the run 3/18
Standards above photometric linearity ceiling
Background control wells behaving as active reactions 1/24
Sample wells with a decreasing trace 0/24
Sample wells with an elevated starting A412 0/24 ()
Samples with ALL THREE replicates baseline-compromised 0/8 ()
Sample wells with a read glitch 0/24
Sample wells usable for rate extraction 24/24
Samples with technical CV > threshold (all reps) 0/8
Samples with technical CV > threshold (usable reps) 0/8
Positive control replicates rising and linear 3/3

Quality control summary for this plate

--- qc_summary: consolidated QC checks for the run ---

'data.frame':   18 obs. of  2 variables:
 $ check: chr  "Occupied wells in layout" "Layout wells missing from absorbance CSV" "Max disagreement between absorbance CSV and full report" "Standard concentrations with replicate CV > threshold" ...
 $ value: chr  "69" "0" "0" "40" ...

10.1 Auto-generated findings

The bullets below are generated programmatically from the QC/CV/anomaly objects above by generate_qc_findings(), so the same logic that flags a problem here is guaranteed to be the logic that flagged it upstream.

qc_findings <- generate_qc_findings(well_diagnostics, well_rates, cv_summary,
                                     baseline_per_sample, std_curve, recon$summary,
                                     assay_params)

cat(paste0("- ", qc_findings, "\n"), sep = "\n")
- **Kinetic sources agree.** absorbance CSV and full report agree exactly on all 759 shared well x timepoint readings; all layout wells are present.

- **Background control wells behaving as active reactions.** 1 of 24 background wells show a starting A412 or drift inconsistent with a flat, inactive well: C8.

- **GSH standard curve replicate imprecision.** 1 of 6 concentrations exceed 15% CV: 40 nmol/well.

- **Standard outlier wells.** 1 of 18 standard wells deviate from their triplicate median by more than 0.15 A412: F5.

- **Standard signal decay.** 3 of 18 standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.

- **Standard curve fit (outlier-excluded):** slope = 0.03075 A412/nmol, R^2 = 0.9985 (all-wells R^2 = 0.9636).

- **Technical replicate CV.** All 8 samples are within 15% CV across all three replicates.

11 SUMMARY

11.1 Anomalies found

  1. One background well behaves as an active reaction, not a background control. C8 (F05_03_36C’s third background replicate) starts at A412 0.252 (above the 0.15 flat-well threshold) and declines to 0.202 (net change -0.050), rather than staying flat. This pattern is consistent with Reaction Mix having been dispensed instead of Background Control Mix for this single well. compute_background_correction() drops it automatically; F05_03_36C’s background estimate is still based on 2 of 3 replicates.
  2. One GSH standard well is an outlier. F5 (40 nmol) reads A412 1.067 against triplicate siblings F4/F6 at 1.433/1.422 — a deviation of -0.355 A412 from the triplicate median, and also flagged for an internal read-to-read glitch. Excluding it, the 40 nmol triplicate CV drops from 15.9% to 0.5%. It is the sole reason the 40 nmol concentration is the only one of six to exceed the 15% CV threshold.
  3. Three standard wells lost signal over the run (F1, F4, F6 - the 32 and 40 nmol replicates; drift -0.004 to -0.011 A412), consistent with TNB2- instability over the run. The standard curve is read at t = 0 for this reason.
  4. No decreasing kinetics, no read glitches, and no elevated starting absorbance in any sample well. All 24 sample wells rise monotonically and cleanly (replicate-level R2 > 0.99), and none needs exclusion from rate extraction.
  5. No sample has an elevated starting absorbance or a disqualifying baseline artifact. Sample well starting A412 ranges only 0.179-0.192, well under the 0.35 threshold.

11.2 Technical replicate precision

All eight samples pass the 15% CV threshold — the highest is F05_08_36C at 7.8%, and every other sample is under 4% (F05_01_36C 3.3%, F05_02_36C 2.4%, F05_03_36C 3.5%, F05_04_36C 2.3%, F05_05_36C 1.2%, F05_06_36C 2.4%, F05_07_36C 1.2%). No replicate is excluded from any sample’s rate calculation on this plate, so the “all reps” and “usable reps only” CVs are identical throughout. No sample technical rep exceeds 15% CV on this plate.

11.3 Assay validity

The positive control worked: all three replicates rise linearly (R2 > 0.998) at 2.8-3.0 mA412/min with low replicate CV. The GSH standard curve has a single, clearly-identified outlier well and every sample well’s kinetics are clean, so this plate’s activity values carry low calibration uncertainty.

11.4 Recommendation

Treat the activity values here as reliable, though the open question about the raw-rate/protein relationship still bears on interpretation:

  1. This plate does not need re-assay. No sample is disqualified, no replicate is excluded, and the standard curve fails QC only by a single, clearly-identified outlier well.
  2. A dilution series is recommended before treating protein-normalized activity as reliable. Raw rate (mU/µL) is essentially flat against total extracted protein here (r = -0.48, p = 0.23), so a division artifact is the more likely explanation for the observed mU/mg-protein trend (see note 3) rather than a genuine effect of protein content on measured activity. A dilution series run on a single homogenate at multiple concentrations would directly test whether the assay’s raw rate is protein-independent.
  3. The mU/mg protein negative trend seen on this plate should be treated as attributable to division artifact rather than a biological effect. Because raw mU/µL is flat against protein while mU/mg protein trends negatively with it, the mU/mg protein figures largely reflect dividing a near-constant numerator by a varying denominator rather than a real difference in specific activity.
  4. Re-confirm the background-well dispensing step, since one well (C8) showed the signature of a Reaction Mix / Background Control Mix mix-up — worth checking whether it is a one-off pipetting error or a persistent low-rate issue in this step.

11.5 What can be used from this plate

All eight samples are usable, and the raw-rate result is the more striking one for the within-family question this project asks: background-corrected rates cluster tightly — 2.13 to 2.30 mA412/min across all eight F05 individuals at 36 °C, a spread of only 7.8% against 1-8% technical CV per sample. Once normalization artifacts are set aside, no striking within-family variation in raw CS activity is apparent at 36 °C.

  • Raw rate is the safer basis for downstream comparison. The median background-corrected rate is ~2.21 mA412/min at 36 °C, with technical and normalization uncertainty small relative to this signal, and it does not require resolving the protein-normalization question first.
  • The mU/mg protein figures on this plate should be treated with some caution. Since raw rate (mU/µL) is essentially flat against total extracted protein here (see note 3 above), the observed mU/mg-protein trend is more consistent with a division artifact than a biological effect of protein content on activity. A dilution series run on a single homogenate at multiple concentrations would help resolve whether the assay’s raw rate is genuinely protein-independent.
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase ABCAM https://robertslab.github.io/sams-notebook/posts/2026/2026-08-14-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-5-36C-Ctenidia/ Fri, 14 Aug 2026 07:00:00 GMT Protein Quantification - June 2026 SORMI M.gigas Ctenidia from Famillies 5 and 7 for Citrate Synthase Assay Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-13-Protein-Quantification---June-2026-SORMI-M.gigas-Ctenidia-from-Famillies-5-and-7-for-Citrate-Synthase-Assay/

INTRO

As part of the June 2026 SORMI experiment citrate synthase assays (GitHub repo), protein quantities were needed for proper normalization of the citrate synthase assays. I quantified protein in ctenidia samples from Families 5 and 7 homozenized on 20260811 (Notebook entry) using the Quick Start Bradford Protein Assay Kit 1 (Bio-Rad; cat: 5000201).

Families 5 & 7 (Eight samples each family from each temperature treatment) were processed for protein quantification.

Due to poor replicates and/or samples falling outside of the standard curve, the BSA assay was repeated for those samples failing QC on the initial run.

MATERIALS & METHODS

BSA Assay

The Bradford Protein Assay was performed according to the manufacturer’s instructions (Bio-Rad, cat: 5000201) for microplate format. Clear, flat-bottomed 96-well plates were used for the assay.

Briefly, samples were thawed, spun at 10,000 rpm for 5 minutes, and stored on ice. A series of BSA standards were prepared and used to generate a standard curve. Samples (5uL) were then added to respective wells and 250uL of the Bradford reagent was added to each well (using a multichannel pipette; two rounds of 125uL) and mixed by pipetting. The plates were then incubated for 5 minutes at room temperature, and the absorbance was measured at 595 nm in a Synergy HTX plate reader (Agilent).

Analysis

Data was analyzed in R. Respective Rmd files are here (GitHub):

Knitted markdown from each is below.

Data files

Plate reader files

Plate reader files (*.xpt)are available here (require Gen5 software to open):

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/BSA/plate_reader_files

Raw absorbance data

Raw absorbance data (*.csv) are available here:

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/BSA/raw_absorbance

RESULTS

Output files with sample concentrations can be found in each of the corresponding output files from the Rmd files above (https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/outputs) in a file named sample_concentrations.csv.

Below is a summary of the protein concentrations (ug/mL) for each sample:

sample_name concentration(ug/mL)
F05_01_36C 953.1
F05_01_ambient 876.1
F05_02_36C 545.8
F05_02_ambient 1038.9
F05_03_36C 667.6
F05_03_ambient 1900.4
F05_04_36C 1062.2
F05_04_ambient 1129.4
F05_05_36C 1666.8
F05_05_ambient 711.7
F05_06_36C 1265.9
F05_06_ambient 2232.4
F05_07_36C 1110.5
F05_07_ambient 1687.8
F05_08_36C 1357.5
F05_08_ambient 2703.7
F07_01_36C 663.3
F07_01_ambient 263.1
F07_02_36C 932.4
F07_02_ambient 1080.1
F07_03_36C 1205.5
F07_03_ambient 1155.6
F07_04_36C 1567.2
F07_04_ambient 560
F07_05_36C 427.9
F07_05_ambient 659.3
F07_06_36C 496.5
F07_06_ambient 1119.4
F07_07_36C 1038.1
F07_07_ambient 1644.6
F07_08_36C 289.8
F07_08_ambient 560.2

AI Use Level 4: Substantial AI contribution

FAMILY 5 ANALYSIS

1 BACKGROUND

Protein quantification (BCA/Bradford-style, 595 nm) of Magallana gigas (Pacific oyster) SoRMI plate F05, 16 samples (F05_01-F05_08, ambient and 36°C exposure) plus an 8-point BSA standard curve, each measured in triplicate wells.

The raw Gen5 export (Gen5-20260813-mgig-BSA-F05-absorbance.csv) is a per-well table where each triplicate group’s first row carries the sample/standard ID and the instrument’s own pre-computed mean/SD/CV-of-concentration, and the two replicate rows below it carry only well-level data. This document ignores the instrument’s pre-computed summary columns and recomputes mean, SD, and CV independently in R via dplyr, for both the raw absorbance (595 nm) and the concentration.

CV QC metric used for exclusion: the >15% technical-replicate CV exclusion rule (including for standard-curve points) is applied to CV of concentration (matching what the instrument’s own CV (%) column already represents for this export – verified against the raw file: e.g. SPL1’s given Mean/SD/CV of 971.656/77.065/7.931 are exactly the mean/SD/CV of its three [Concentration] values, not of its 595 values). Concentration CV is the more sensitive metric here: several groups exceed 15% on concentration while none do on raw absorbance, since back-calculation against a curve with a non-zero intercept amplifies small absorbance differences into larger relative concentration differences.

Two edge cases in the raw [Concentration] column need explicit handling:

  • Censored reads. Wells reading outside the instrument’s own internal standard-curve range are reported as <0.000 or >2100.000 rather than a number. These are parsed to a _censored flag plus the boundary value, and excluded from concentration mean/SD/CV calculations (their true value is unknown, not equal to the boundary).
  • The zero (blank) BSA standard (STD8) is always censored (<0.000 on all three wells), by construction – a true-zero standard’s back-calculated concentration is not a meaningful quantity to compute a CV on, and this is not a data-quality problem. STD8 is therefore exempt from the concentration-CV exclusion rule and is retained as the curve’s origin point on the strength of its (excellent) 595 CV instead.

Any other group left with fewer than 2 valid (non-censored) concentration replicates – e.g. SPL16 / F05_08_ambient, where 2 of 3 wells read above the top standard – has an undefined (NA) concentration CV and is treated as failing QC, consistent with the source file itself being unable to compute a CV there either (?????).

2 SETUP

2.1 Libraries

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Set output directory

output_dir <- "../outputs/Gen5-20260813-mgig-BSA-F05"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

3 DATA IMPORT

The raw file is a Gen5 per-well export with non-syntactic column names ([Concentration], CV (%), etc.) and a UTF-8 byte-order mark on the first header cell, so columns are renamed by position immediately after import rather than referenced by their original names.

data_path <- "../data/BSA/raw_absorbance/Gen5-20260813-mgig-BSA-F05-absorbance.csv"

protein_raw <- read.csv(data_path, stringsAsFactors = FALSE, check.names = FALSE,
                         na.strings = c("", "NA"))
colnames(protein_raw) <- c("well_id", "name", "well", "std_conc_known", "abs_595",
                           "gen5_concentration", "gen5_count", "gen5_mean",
                           "gen5_sd", "gen5_cv")

cat("--- protein_raw: as imported from the Gen5 export ---\n")
str(protein_raw)
--- protein_raw: as imported from the Gen5 export ---
'data.frame':   72 obs. of  10 variables:
 $ well_id           : chr  "SPL1" NA NA "SPL2" ...
 $ name              : chr  "F05_01_36C" NA NA "F05_02_36C" ...
 $ well              : chr  "A1" "A2" "A3" "A4" ...
 $ std_conc_known    : int  NA NA NA NA NA NA NA NA NA NA ...
 $ abs_595           : num  1.197 1.147 1.123 0.926 0.973 ...
 $ gen5_concentration: chr  "1056.125" "953.667" "905.177" "506.911" ...
 $ gen5_count        : int  3 NA NA 3 NA NA 3 NA NA 3 ...
 $ gen5_mean         : chr  "971.656" NA NA "579.274" ...
 $ gen5_sd           : chr  "77.065" NA NA "64.041" ...
 $ gen5_cv           : chr  "7.931" NA NA "11.055" ...

4 RESHAPING

well_id and name are only populated on the first row of each triplicate group in the raw export, so they are filled downward. Each row is also classified as a standard (well_id starts with STD) or a sample (SPL), and the instrument’s censored concentration strings (<0.000, >2100.000) are parsed into a numeric value plus a gen5_concentration_censored flag.

protein_long <- protein_raw %>%
  select(well_id, name, well, std_conc_known, abs_595, gen5_concentration) %>%
  fill(well_id, name, .direction = "down") %>%
  mutate(
    type = if_else(grepl("^STD", well_id), "standard", "sample"),
    gen5_concentration_censored = grepl("^[<>]", gen5_concentration),
    gen5_concentration_numeric  = as.numeric(gsub("[<>]", "", gen5_concentration))
  )

cat("--- protein_long: one row per well, group identifiers filled down ---\n")
str(protein_long)
--- protein_long: one row per well, group identifiers filled down ---
'data.frame':   72 obs. of  9 variables:
 $ well_id                    : chr  "SPL1" "SPL1" "SPL1" "SPL2" ...
 $ name                       : chr  "F05_01_36C" "F05_01_36C" "F05_01_36C" "F05_02_36C" ...
 $ well                       : chr  "A1" "A2" "A3" "A4" ...
 $ std_conc_known             : int  NA NA NA NA NA NA NA NA NA NA ...
 $ abs_595                    : num  1.197 1.147 1.123 0.926 0.973 ...
 $ gen5_concentration         : chr  "1056.125" "953.667" "905.177" "506.911" ...
 $ type                       : chr  "sample" "sample" "sample" "sample" ...
 $ gen5_concentration_censored: logi  FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ gen5_concentration_numeric : num  1056 954 905 507 602 ...

5 MEAN VALUES

5.1 Mean 595 nm absorbance

mean_595_by_group <- protein_long %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(n_wells  = n(),
            mean_595 = mean(abs_595),
            .groups  = "drop")

cat("--- mean_595_by_group ---\n")
str(mean_595_by_group)
--- mean_595_by_group ---
tibble [24 × 6] (S3: tbl_df/tbl/data.frame)
 $ well_id       : chr [1:24] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name          : chr [1:24] "F05_01_36C" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ type          : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ std_conc_known: int [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells       : int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595      : num [1:24] 1.16 1.21 1.56 1.24 1.04 ...

5.2 Mean concentration

Censored wells (reading outside the instrument’s own standard-curve range) are dropped before averaging, since their true concentration is unknown.

mean_conc_by_group <- protein_long %>%
  filter(!gen5_concentration_censored) %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(n_wells_valid_conc  = n(),
            mean_concentration  = mean(gen5_concentration_numeric),
            .groups = "drop")

cat("--- mean_conc_by_group ---\n")
str(mean_conc_by_group)
--- mean_conc_by_group ---
tibble [23 × 6] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:23] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name              : chr [1:23] "F05_01_36C" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ type              : chr [1:23] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:23] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells_valid_conc: int [1:23] 3 3 3 3 3 3 3 1 3 3 ...
 $ mean_concentration: num [1:23] 972 1073 1796 1143 739 ...

6 STANDARD DEVIATION AND COEFFICIENT OF VARIATION

6.1 SD and CV of 595 nm absorbance

sd_cv_595_by_group <- protein_long %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(
    n_wells  = n(),
    mean_595 = mean(abs_595),
    sd_595   = sd(abs_595),
    cv_595   = 100 * sd_595 / mean_595,
    .groups  = "drop"
  )

cat("--- sd_cv_595_by_group ---\n")
str(sd_cv_595_by_group)
--- sd_cv_595_by_group ---
tibble [24 × 8] (S3: tbl_df/tbl/data.frame)
 $ well_id       : chr [1:24] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name          : chr [1:24] "F05_01_36C" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ type          : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ std_conc_known: int [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells       : int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595      : num [1:24] 1.16 1.21 1.56 1.24 1.04 ...
 $ sd_595        : num [1:24] 0.0378 0.1275 0.0528 0.0188 0.0146 ...
 $ cv_595        : num [1:24] 3.27 10.58 3.38 1.52 1.4 ...

6.2 SD and CV of concentration

sd_cv_conc_by_group <- protein_long %>%
  filter(!gen5_concentration_censored) %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(
    n_wells_valid_conc  = n(),
    mean_concentration  = mean(gen5_concentration_numeric),
    sd_concentration    = sd(gen5_concentration_numeric),
    cv_concentration    = 100 * sd_concentration / mean_concentration,
    .groups = "drop"
  )

cat("--- sd_cv_conc_by_group ---\n")
str(sd_cv_conc_by_group)
--- sd_cv_conc_by_group ---
tibble [23 × 8] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:23] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name              : chr [1:23] "F05_01_36C" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ type              : chr [1:23] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:23] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells_valid_conc: int [1:23] 3 3 3 3 3 3 3 1 3 3 ...
 $ mean_concentration: num [1:23] 972 1073 1796 1143 739 ...
 $ sd_concentration  : num [1:23] 77.1 258.7 107.8 38.2 30.1 ...
 $ cv_concentration  : num [1:23] 7.93 24.1 6 3.34 4.07 ...

6.3 Combined per-group summary

group_stats <- sd_cv_595_by_group %>%
  left_join(
    sd_cv_conc_by_group %>%
      select(well_id, mean_concentration, sd_concentration, cv_concentration, n_wells_valid_conc),
    by = "well_id"
  ) %>%
  arrange(type, well_id)

cat("--- group_stats: one row per sample/standard, 595- and concentration-level stats ---\n")
str(group_stats)
kable(group_stats, digits = 3)
--- group_stats: one row per sample/standard, 595- and concentration-level stats ---
tibble [24 × 12] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:24] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name              : chr [1:24] "F05_01_36C" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ type              : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells           : int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595          : num [1:24] 1.16 1.21 1.56 1.24 1.04 ...
 $ sd_595            : num [1:24] 0.0378 0.1275 0.0528 0.0188 0.0146 ...
 $ cv_595            : num [1:24] 3.27 10.58 3.38 1.52 1.4 ...
 $ mean_concentration: num [1:24] 972 1073 1796 1143 739 ...
 $ sd_concentration  : num [1:24] 77.1 258.7 107.8 38.2 30.1 ...
 $ cv_concentration  : num [1:24] 7.93 24.1 6 3.34 4.07 ...
 $ n_wells_valid_conc: int [1:24] 3 3 3 3 3 3 3 1 3 3 ...
well_id name type std_conc_known n_wells mean_595 sd_595 cv_595 mean_concentration sd_concentration cv_concentration n_wells_valid_conc
SPL1 F05_01_36C sample NA 3 1.156 0.038 3.267 971.656 77.065 7.931 3
SPL10 F05_02_ambient sample NA 3 1.206 0.128 10.575 1073.167 258.654 24.102 3
SPL11 F05_03_ambient sample NA 3 1.562 0.053 3.379 1795.714 107.767 6.001 3
SPL12 F05_04_ambient sample NA 3 1.240 0.019 1.518 1142.757 38.212 3.344 3
SPL13 F05_05_ambient sample NA 3 1.041 0.015 1.400 738.878 30.098 4.074 3
SPL14 F05_06_ambient sample NA 3 1.626 0.060 3.684 1925.899 121.472 6.307 3
SPL15 F05_07_ambient sample NA 3 1.506 0.076 5.036 1682.300 154.559 9.187 3
SPL16 F05_08_ambient sample NA 3 1.719 0.040 2.336 2030.183 NA NA 1
SPL2 F05_02_36C sample NA 3 0.962 0.032 3.282 579.274 64.041 11.055 3
SPL3 F05_03_36C sample NA 3 1.020 0.025 2.411 696.136 49.535 7.116 3
SPL4 F05_04_36C sample NA 3 1.208 0.008 0.643 1077.834 15.456 1.434 3
SPL5 F05_05_36C sample NA 3 1.496 0.075 5.025 1661.606 153.165 9.218 3
SPL6 F05_06_36C sample NA 3 1.305 0.060 4.575 1274.228 121.807 9.559 3
SPL7 F-05_07_36C sample NA 3 1.231 0.064 5.218 1123.956 129.443 11.517 3
SPL8 F05_08_36C sample NA 3 1.348 0.054 4.022 1362.619 110.153 8.084 3
SPL9 F05_01_ambient sample NA 3 1.063 0.067 6.331 782.972 136.834 17.476 3
STD1 BSA standard 2000 3 1.560 0.059 3.802 1792.265 120.228 6.708 3
STD2 BSA standard 1500 3 1.411 0.035 2.451 1490.640 69.690 4.675 3
STD3 BSA standard 1000 3 1.299 0.025 1.887 1262.866 49.516 3.921 3
STD4 BSA standard 750 3 1.145 0.055 4.809 951.300 112.197 11.794 3
STD5 BSA standard 500 3 0.960 0.025 2.551 575.554 49.611 8.620 3
STD6 BSA standard 250 3 0.768 0.029 3.794 185.471 58.758 31.680 3
STD7 BSA standard 125 3 0.715 0.010 1.402 77.197 21.167 27.419 3
STD8 BSA standard 0 3 0.573 0.003 0.524 NA NA NA NA

7 CV-BASED EXCLUSION

Groups with CV(concentration) > 15% – or an undefined concentration CV due to insufficient valid (non-censored) replicates – are flagged and excluded from all downstream analysis, including standard-curve fitting. The zero BSA standard (STD8) is exempt from this rule (see Background) and is always retained.

group_stats <- group_stats %>%
  mutate(
    is_zero_standard = type == "standard" & !is.na(std_conc_known) & std_conc_known == 0,
    high_cv = !is_zero_standard & (is.na(cv_concentration) | cv_concentration > 15)
  )

excluded_groups <- group_stats %>% filter(high_cv)

cat("--- excluded_groups: CV(concentration) > 15% (or undefined), dropped from further analysis ---\n")
str(excluded_groups)
kable(excluded_groups %>%
        select(well_id, name, type, std_conc_known, n_wells_valid_conc,
               mean_concentration, sd_concentration, cv_concentration),
      digits = 3)

clean_group_stats <- group_stats %>% filter(!high_cv)

cat("\nRetained after CV exclusion:", nrow(clean_group_stats), "of", nrow(group_stats), "groups\n")
str(clean_group_stats)
--- excluded_groups: CV(concentration) > 15% (or undefined), dropped from further analysis ---
tibble [5 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:5] "SPL10" "SPL16" "SPL9" "STD6" ...
 $ name              : chr [1:5] "F05_02_ambient" "F05_08_ambient" "F05_01_ambient" "BSA" ...
 $ type              : chr [1:5] "sample" "sample" "sample" "standard" ...
 $ std_conc_known    : int [1:5] NA NA NA 250 125
 $ n_wells           : int [1:5] 3 3 3 3 3
 $ mean_595          : num [1:5] 1.206 1.719 1.063 0.768 0.715
 $ sd_595            : num [1:5] 0.1275 0.0401 0.0673 0.0291 0.01
 $ cv_595            : num [1:5] 10.58 2.34 6.33 3.79 1.4
 $ mean_concentration: num [1:5] 1073.2 2030.2 783 185.5 77.2
 $ sd_concentration  : num [1:5] 258.7 NA 136.8 58.8 21.2
 $ cv_concentration  : num [1:5] 24.1 NA 17.5 31.7 27.4
 $ n_wells_valid_conc: int [1:5] 3 1 3 3 3
 $ is_zero_standard  : logi [1:5] FALSE FALSE FALSE FALSE FALSE
 $ high_cv           : logi [1:5] TRUE TRUE TRUE TRUE TRUE
well_id name type std_conc_known n_wells_valid_conc mean_concentration sd_concentration cv_concentration
SPL10 F05_02_ambient sample NA 3 1073.167 258.654 24.102
SPL16 F05_08_ambient sample NA 1 2030.183 NA NA
SPL9 F05_01_ambient sample NA 3 782.972 136.834 17.476
STD6 BSA standard 250 3 185.471 58.758 31.680
STD7 BSA standard 125 3 77.197 21.167 27.419 
Retained after CV exclusion: 19 of 24 groups
tibble [19 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:19] "SPL1" "SPL11" "SPL12" "SPL13" ...
 $ name              : chr [1:19] "F05_01_36C" "F05_03_ambient" "F05_04_ambient" "F05_05_ambient" ...
 $ type              : chr [1:19] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:19] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells           : int [1:19] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595          : num [1:19] 1.16 1.56 1.24 1.04 1.63 ...
 $ sd_595            : num [1:19] 0.0378 0.0528 0.0188 0.0146 0.0599 ...
 $ cv_595            : num [1:19] 3.27 3.38 1.52 1.4 3.68 ...
 $ mean_concentration: num [1:19] 972 1796 1143 739 1926 ...
 $ sd_concentration  : num [1:19] 77.1 107.8 38.2 30.1 121.5 ...
 $ cv_concentration  : num [1:19] 7.93 6 3.34 4.07 6.31 ...
 $ n_wells_valid_conc: int [1:19] 3 3 3 3 3 3 3 3 3 3 ...
 $ is_zero_standard  : logi [1:19] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv           : logi [1:19] FALSE FALSE FALSE FALSE FALSE FALSE ...

8 STANDARD CURVE

Fit using only the retained (CV ≤ 15%) standard points: known BSA concentration (Conc/Dil) on the x-axis, mean 595 nm absorbance on the y-axis.

standards_clean <- clean_group_stats %>% filter(type == "standard")

cat("--- standards_clean: standard points retained for curve fitting ---\n")
str(standards_clean)

standard_curve_model <- lm(mean_595 ~ std_conc_known, data = standards_clean)

curve_slope     <- unname(coef(standard_curve_model)[2])
curve_intercept <- unname(coef(standard_curve_model)[1])
curve_r2        <- summary(standard_curve_model)$r.squared

cat("Standard curve fit: 595 =", format(curve_slope, digits = 6), "* concentration +",
    format(curve_intercept, digits = 4), "\n")
cat("R2:", round(curve_r2, 5), "\n")
--- standards_clean: standard points retained for curve fitting ---
tibble [6 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:6] "STD1" "STD2" "STD3" "STD4" ...
 $ name              : chr [1:6] "BSA" "BSA" "BSA" "BSA" ...
 $ type              : chr [1:6] "standard" "standard" "standard" "standard" ...
 $ std_conc_known    : int [1:6] 2000 1500 1000 750 500 0
 $ n_wells           : int [1:6] 3 3 3 3 3 3
 $ mean_595          : num [1:6] 1.56 1.41 1.3 1.15 0.96 ...
 $ sd_595            : num [1:6] 0.0593 0.0346 0.0245 0.0551 0.0245 ...
 $ cv_595            : num [1:6] 3.8 2.45 1.89 4.81 2.55 ...
 $ mean_concentration: num [1:6] 1792 1491 1263 951 576 ...
 $ sd_concentration  : num [1:6] 120.2 69.7 49.5 112.2 49.6 ...
 $ cv_concentration  : num [1:6] 6.71 4.68 3.92 11.79 8.62 ...
 $ n_wells_valid_conc: int [1:6] 3 3 3 3 3 NA
 $ is_zero_standard  : logi [1:6] FALSE FALSE FALSE FALSE FALSE TRUE
 $ high_cv           : logi [1:6] FALSE FALSE FALSE FALSE FALSE FALSE
Standard curve fit: 595 = 0.000476359 * concentration + 0.7017 
R2: 0.92345

8.1 Plot standard curve

eq_label <- paste0("y = ", format(curve_slope, digits = 4, scientific = TRUE), "x + ",
                    sprintf("%.4f", curve_intercept),
                    "\nR² = ", sprintf("%.4f", curve_r2))

standard_curve_plot <- ggplot(standards_clean, aes(x = std_conc_known, y = mean_595)) +
  geom_smooth(method = "lm", formula = y ~ x, se = FALSE, color = "steelblue", linewidth = 0.6) +
  geom_errorbar(aes(ymin = mean_595 - sd_595, ymax = mean_595 + sd_595),
                width = 30, color = "grey40") +
  geom_point(size = 2.6, color = "steelblue") +
  annotate("text", x = -Inf, y = Inf, label = eq_label,
           hjust = -0.08, vjust = 1.3, size = 3.6, fontface = "bold") +
  labs(x = "BSA standard concentration (µg/mL)",
       y = "Mean absorbance (595 nm)",
       title = "BSA standard curve",
       subtitle = "Points with technical-replicate CV(concentration) > 15% excluded (see below)") +
  theme_bw(base_size = 12)

print(standard_curve_plot)

ggsave(file.path(output_dir, "standard_curve_595.png"), standard_curve_plot,
       width = 7, height = 5.5, dpi = 300)

Standard points excluded from the curve:

excluded_standards <- group_stats %>% filter(type == "standard", high_cv)

cat("--- excluded_standards ---\n")
str(excluded_standards)
kable(excluded_standards %>%
        select(well_id, std_conc_known, mean_concentration, sd_concentration, cv_concentration),
      digits = 3)
--- excluded_standards ---
tibble [2 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:2] "STD6" "STD7"
 $ name              : chr [1:2] "BSA" "BSA"
 $ type              : chr [1:2] "standard" "standard"
 $ std_conc_known    : int [1:2] 250 125
 $ n_wells           : int [1:2] 3 3
 $ mean_595          : num [1:2] 0.768 0.715
 $ sd_595            : num [1:2] 0.0291 0.01
 $ cv_595            : num [1:2] 3.79 1.4
 $ mean_concentration: num [1:2] 185.5 77.2
 $ sd_concentration  : num [1:2] 58.8 21.2
 $ cv_concentration  : num [1:2] 31.7 27.4
 $ n_wells_valid_conc: int [1:2] 3 3
 $ is_zero_standard  : logi [1:2] FALSE FALSE
 $ high_cv           : logi [1:2] TRUE TRUE
well_id std_conc_known mean_concentration sd_concentration cv_concentration
STD6 250 185.471 58.758 31.680
STD7 125 77.197 21.167 27.419

9 APPLYING THE STANDARD CURVE TO SAMPLES

Samples that passed CV QC are back-calculated against the fitted curve. Samples whose mean 595 falls outside the retained standards’ absorbance range are flagged as extrapolated (e.g. F05_08_ambient read above the top standard on 2 of its 3 wells and was excluded above on CV grounds already, since only 1 of its wells had a defined, non-censored concentration).

samples_clean <- clean_group_stats %>%
  filter(type == "sample") %>%
  mutate(
    se_595                    = sd_595 / sqrt(n_wells),
    calculated_concentration  = (mean_595 - curve_intercept) / curve_slope,
    in_curve_range            = mean_595 >= min(standards_clean$mean_595) &
                                  mean_595 <= max(standards_clean$mean_595)
  ) %>%
  arrange(name)

cat("--- samples_clean: CV-QC-passing samples, quantified against the fitted standard curve ---\n")
str(samples_clean)
--- samples_clean: CV-QC-passing samples, quantified against the fitted standard curve ---
tibble [13 × 17] (S3: tbl_df/tbl/data.frame)
 $ well_id                 : chr [1:13] "SPL7" "SPL1" "SPL2" "SPL3" ...
 $ name                    : chr [1:13] "F-05_07_36C" "F05_01_36C" "F05_02_36C" "F05_03_36C" ...
 $ type                    : chr [1:13] "sample" "sample" "sample" "sample" ...
 $ std_conc_known          : int [1:13] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells                 : int [1:13] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595                : num [1:13] 1.231 1.156 0.962 1.02 1.562 ...
 $ sd_595                  : num [1:13] 0.0642 0.0378 0.0316 0.0246 0.0528 ...
 $ cv_595                  : num [1:13] 5.22 3.27 3.28 2.41 3.38 ...
 $ mean_concentration      : num [1:13] 1124 972 579 696 1796 ...
 $ sd_concentration        : num [1:13] 129.4 77.1 64 49.5 107.8 ...
 $ cv_concentration        : num [1:13] 11.52 7.93 11.06 7.12 6 ...
 $ n_wells_valid_conc      : int [1:13] 3 3 3 3 3 3 3 3 3 3 ...
 $ is_zero_standard        : logi [1:13] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv                 : logi [1:13] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ se_595                  : num [1:13] 0.0371 0.0218 0.0182 0.0142 0.0305 ...
 $ calculated_concentration: num [1:13] 1111 953 546 668 1805 ...
 $ in_curve_range          : logi [1:13] TRUE TRUE TRUE TRUE FALSE TRUE ...
cat("Samples with absorbance outside the retained standard curve's range (extrapolated):\n")
kable(samples_clean %>% filter(!in_curve_range) %>%
        select(well_id, name, mean_595, calculated_concentration),
      digits = 3)
Samples with absorbance outside the retained standard curve's range (extrapolated):
well_id name mean_595 calculated_concentration
SPL11 F05_03_ambient 1.562 1805.383
SPL14 F05_06_ambient 1.626 1940.435

9.1 Plot samples on standard curve

Every QC-passing sample is plotted at its own mean 595 and back-calculated concentration, with vertical error bars of ±1 standard error (of mean 595, across its 2-3 retained technical replicates), overlaid on the standard curve fit line and retained standard points.

samples_on_curve_plot <- ggplot() +
  geom_smooth(data = standards_clean, aes(x = std_conc_known, y = mean_595),
              method = "lm", formula = y ~ x, se = FALSE,
              color = "steelblue", linewidth = 0.6) +
  geom_point(data = standards_clean, aes(x = std_conc_known, y = mean_595),
             size = 2.6, color = "steelblue") +
  geom_errorbar(data = samples_clean,
                aes(x = calculated_concentration,
                    ymin = mean_595 - se_595, ymax = mean_595 + se_595),
                width = 30, color = "grey40") +
  geom_point(data = samples_clean,
             aes(x = calculated_concentration, y = mean_595, shape = in_curve_range),
             size = 2.4, color = "firebrick", alpha = 0.85) +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                      labels = c(`TRUE` = "in range", `FALSE` = "extrapolated"),
                      name = "Sample status") +
  labs(x = "Concentration (µg/mL)",
       y = "Mean absorbance (595 nm)",
       title = "BSA standard curve with QC-passing samples overlaid",
       subtitle = "Samples plotted at their back-calculated concentration; error bars = ±1 SE of mean 595") +
  theme_bw(base_size = 12)

print(samples_on_curve_plot)

ggsave(file.path(output_dir, "standard_curve_with_samples.png"), samples_on_curve_plot,
       width = 8, height = 5.5, dpi = 300)

10 RESULTS TABLE

results_table <- samples_clean %>%
  transmute(
    Sample                              = name,
    `Well ID`                           = well_id,
    `Mean 595`                          = round(mean_595, 3),
    `CV 595 (%)`                        = round(cv_595, 2),
    `CV concentration (%)`              = round(cv_concentration, 2),
    `Calculated concentration (ug/mL)`  = round(calculated_concentration, 1),
    `In standard curve range`           = ifelse(in_curve_range, "yes", "NO - extrapolated")
  ) %>%
  arrange(gsub("[0-9]+$", "", `Well ID`), as.numeric(gsub("^[A-Za-z]+", "", `Well ID`)))

kable(results_table)
Sample Well ID Mean 595 CV 595 (%) CV concentration (%) Calculated concentration (ug/mL) In standard curve range
F05_01_36C SPL1 1.156 3.27 7.93 953.1 yes
F05_02_36C SPL2 0.962 3.28 11.06 545.8 yes
F05_03_36C SPL3 1.020 2.41 7.12 667.6 yes
F05_04_36C SPL4 1.208 0.64 1.43 1062.2 yes
F05_05_36C SPL5 1.496 5.03 9.22 1666.8 yes
F05_06_36C SPL6 1.305 4.57 9.56 1265.9 yes
F-05_07_36C SPL7 1.231 5.22 11.52 1110.5 yes
F05_08_36C SPL8 1.348 4.02 8.08 1357.5 yes
F05_03_ambient SPL11 1.562 3.38 6.00 1805.4 NO - extrapolated
F05_04_ambient SPL12 1.240 1.52 3.34 1129.4 yes
F05_05_ambient SPL13 1.041 1.40 4.07 711.7 yes
F05_06_ambient SPL14 1.626 3.68 6.31 1940.4 NO - extrapolated
F05_07_ambient SPL15 1.506 5.04 9.19 1687.8 yes

Only samples passing all QC – CV ≤ 15% and within the retained standard curve’s range (i.e. not extrapolated) – are written to the CSV output.

results_table_export <- results_table %>% filter(`In standard curve range` == "yes")

cat("--- results_table_export: samples passing all QC (CV and in-range), written to CSV ---\n")
str(results_table_export)

write.csv(results_table_export, file.path(output_dir, "sample_protein_concentrations.csv"), row.names = FALSE)
--- results_table_export: samples passing all QC (CV and in-range), written to CSV ---
tibble [11 × 7] (S3: tbl_df/tbl/data.frame)
 $ Sample                          : chr [1:11] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_04_36C" ...
 $ Well ID                         : chr [1:11] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ Mean 595                        : num [1:11] 1.156 0.962 1.02 1.208 1.496 ...
 $ CV 595 (%)                      : num [1:11] 3.27 3.28 2.41 0.64 5.03 4.57 5.22 4.02 1.52 1.4 ...
 $ CV concentration (%)            : num [1:11] 7.93 11.06 7.12 1.43 9.22 ...
 $ Calculated concentration (ug/mL): num [1:11] 953 546 668 1062 1667 ...
 $ In standard curve range         : chr [1:11] "yes" "yes" "yes" "yes" ...

11 SUMMARY

  • 24 sample/standard triplicate groups were imported (8 BSA standards, 16 samples).
  • 5 group(s) exceeded 15% technical-replicate CV(concentration), or had an undefined CV from insufficient valid replicates, and were excluded from all downstream analysis: SPL10 (F05_02_ambient), SPL16 (F05_08_ambient), SPL9 (F05_01_ambient), STD6 (BSA), STD7 (BSA).
  • The BSA standard curve was fit on the 6 remaining standard points: slope = 0.0004764, intercept = 0.7017, R² = 0.9234.
  • 2 CV-QC-passing sample(s) fell outside the retained standard curve’s absorbance range and are flagged as extrapolated in the results table above.

11.1 Samples passing QC

Sample names are sorted naturally (e.g. F05_02_ambient before F05_10_ambient, and ignoring stray punctuation like the F-05_07_36C typo) rather than by plain lexicographic string order.

natural_sort_key <- function(x) gsub("[^A-Za-z0-9]", "", x)
passing_samples_table <- samples_clean %>%
  transmute(sample_name = name,
            `concentration(ug/mL)` = round(calculated_concentration, 1)) %>%
  arrange(natural_sort_key(sample_name))

cat("--- passing_samples_table: samples passing CV QC, name + calculated concentration ---\n")
str(passing_samples_table)

kable(passing_samples_table)
--- passing_samples_table: samples passing CV QC, name + calculated concentration ---
tibble [13 × 2] (S3: tbl_df/tbl/data.frame)
 $ sample_name         : chr [1:13] "F05_01_36C" "F05_02_36C" "F05_03_36C" "F05_03_ambient" ...
 $ concentration(ug/mL): num [1:13] 953 546 668 1805 1062 ...
sample_name concentration(ug/mL)
F05_01_36C 953.1
F05_02_36C 545.8
F05_03_36C 667.6
F05_03_ambient 1805.4
F05_04_36C 1062.2
F05_04_ambient 1129.4
F05_05_36C 1666.8
F05_05_ambient 711.7
F05_06_36C 1265.9
F05_06_ambient 1940.4
F-05_07_36C 1110.5
F05_07_ambient 1687.8
F05_08_36C 1357.5

FAMILY 7 ANALYSIS

1 BACKGROUND

Protein quantification (BCA/Bradford-style, 595 nm) of Magallana gigas (Pacific oyster) SoRMI plate F07, 16 samples (F07_01-F07_08, ambient and 36°C exposure) plus an 8-point BSA standard curve, each measured in triplicate wells.

The raw Gen5 export (Gen5-20260813-mgig-BSA-F07-absorbance.csv) is a per-well table where each triplicate group’s first row carries the sample/standard ID and the instrument’s own pre-computed mean/SD/CV-of-concentration, and the two replicate rows below it carry only well-level data. This document ignores the instrument’s pre-computed summary columns and recomputes mean, SD, and CV independently in R via dplyr, for both the raw absorbance (595 nm) and the concentration.

CV QC metric used for exclusion: the >15% technical-replicate CV exclusion rule (including for standard-curve points) is applied to CV of concentration (matching what the instrument’s own CV (%) column already represents for this export – verified against the raw file: e.g. SPL1’s given Mean/SD/CV of 694.467/49.577/7.139 are exactly the mean/SD/CV of its three [Concentration] values, not of its 595 values). Concentration CV is the more sensitive metric here: several groups exceed 15% on concentration while none do on raw absorbance, since back-calculation against a curve with a non-zero intercept amplifies small absorbance differences into larger relative concentration differences.

Two edge cases in the raw [Concentration] column need explicit handling:

  • Censored reads. Wells reading outside the instrument’s own internal standard-curve range are reported as <0.000 or >2100.000 rather than a number. These are parsed to a _censored flag plus the boundary value, and excluded from concentration mean/SD/CV calculations (their true value is unknown, not equal to the boundary).
  • The zero (blank) BSA standard (STD8) is always censored (<0.000 on all three wells), by construction – a true-zero standard’s back-calculated concentration is not a meaningful quantity to compute a CV on, and this is not a data-quality problem. STD8 is therefore exempt from the concentration-CV exclusion rule and is retained as the curve’s origin point on the strength of its (excellent) 595 CV instead.

Any other group left with fewer than 2 valid (non-censored) concentration replicates has an undefined (NA) concentration CV and is treated as failing QC, consistent with the source file itself being unable to compute a CV there either (?????). In this plate’s data, every sample well yields a numeric (non-censored) concentration, so this case only arises for the zero BSA standard (which is exempted, above).

2 SETUP

2.1 Libraries

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Set output directory

output_dir <- "../outputs/Gen5-20260813-mgig-BSA-F07"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

3 DATA IMPORT

The raw file is a Gen5 per-well export with non-syntactic column names ([Concentration], CV (%), etc.) and a UTF-8 byte-order mark on the first header cell, so columns are renamed by position immediately after import rather than referenced by their original names.

data_path <- "../data/BSA/raw_absorbance/Gen5-20260813-mgig-BSA-F07-absorbance.csv"

protein_raw <- read.csv(data_path, stringsAsFactors = FALSE, check.names = FALSE,
                         na.strings = c("", "NA"))
colnames(protein_raw) <- c("well_id", "name", "well", "std_conc_known", "abs_595",
                           "gen5_concentration", "gen5_count", "gen5_mean",
                           "gen5_sd", "gen5_cv")

cat("--- protein_raw: as imported from the Gen5 export ---\n")
str(protein_raw)
--- protein_raw: as imported from the Gen5 export ---
'data.frame':   72 obs. of  10 variables:
 $ well_id           : chr  "SPL1" NA NA "SPL2" ...
 $ name              : chr  "F07_01_36C" NA NA "F07_02_36C" ...
 $ well              : chr  "A1" "A2" "A3" "A4" ...
 $ std_conc_known    : int  NA NA NA NA NA NA NA NA NA NA ...
 $ abs_595           : num  1.01 1.04 1.07 1.15 1.15 ...
 $ gen5_concentration: chr  "645.965" "692.383" "745.053" "909.883" ...
 $ gen5_count        : int  3 NA NA 3 NA NA 3 NA NA 3 ...
 $ gen5_mean         : chr  "694.467" NA NA "953.017" ...
 $ gen5_sd           : chr  "49.577" NA NA "81.025" ...
 $ gen5_cv           : chr  "7.139" NA NA "8.502" ...

4 RESHAPING

well_id and name are only populated on the first row of each triplicate group in the raw export, so they are filled downward. Each row is also classified as a standard (well_id starts with STD) or a sample (SPL), and the instrument’s censored concentration strings (<0.000, >2100.000) are parsed into a numeric value plus a gen5_concentration_censored flag.

protein_long <- protein_raw %>%
  select(well_id, name, well, std_conc_known, abs_595, gen5_concentration) %>%
  fill(well_id, name, .direction = "down") %>%
  mutate(
    type = if_else(grepl("^STD", well_id), "standard", "sample"),
    gen5_concentration_censored = grepl("^[<>]", gen5_concentration),
    gen5_concentration_numeric  = as.numeric(gsub("[<>]", "", gen5_concentration))
  )

cat("--- protein_long: one row per well, group identifiers filled down ---\n")
str(protein_long)
--- protein_long: one row per well, group identifiers filled down ---
'data.frame':   72 obs. of  9 variables:
 $ well_id                    : chr  "SPL1" "SPL1" "SPL1" "SPL2" ...
 $ name                       : chr  "F07_01_36C" "F07_01_36C" "F07_01_36C" "F07_02_36C" ...
 $ well                       : chr  "A1" "A2" "A3" "A4" ...
 $ std_conc_known             : int  NA NA NA NA NA NA NA NA NA NA ...
 $ abs_595                    : num  1.01 1.04 1.07 1.15 1.15 ...
 $ gen5_concentration         : chr  "645.965" "692.383" "745.053" "909.883" ...
 $ type                       : chr  "sample" "sample" "sample" "sample" ...
 $ gen5_concentration_censored: logi  FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ gen5_concentration_numeric : num  646 692 745 910 903 ...

5 MEAN VALUES

5.1 Mean 595 nm absorbance

mean_595_by_group <- protein_long %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(n_wells  = n(),
            mean_595 = mean(abs_595),
            .groups  = "drop")

cat("--- mean_595_by_group ---\n")
str(mean_595_by_group)
--- mean_595_by_group ---
tibble [24 × 6] (S3: tbl_df/tbl/data.frame)
 $ well_id       : chr [1:24] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name          : chr [1:24] "F07_01_36C" "F07_02_ambient" "F07_03_ambient" "F07_04_ambient" ...
 $ type          : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ std_conc_known: int [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells       : int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595      : num [1:24] 1.04 1.25 1.29 1.02 1.04 ...

5.2 Mean concentration

Censored wells (reading outside the instrument’s own standard-curve range) are dropped before averaging, since their true concentration is unknown.

mean_conc_by_group <- protein_long %>%
  filter(!gen5_concentration_censored) %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(n_wells_valid_conc  = n(),
            mean_concentration  = mean(gen5_concentration_numeric),
            .groups = "drop")

cat("--- mean_conc_by_group ---\n")
str(mean_conc_by_group)
--- mean_conc_by_group ---
tibble [23 × 6] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:23] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name              : chr [1:23] "F07_01_36C" "F07_02_ambient" "F07_03_ambient" "F07_04_ambient" ...
 $ type              : chr [1:23] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:23] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells_valid_conc: int [1:23] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_concentration: num [1:23] 694 1096 1168 652 691 ...

6 STANDARD DEVIATION AND COEFFICIENT OF VARIATION

6.1 SD and CV of 595 nm absorbance

sd_cv_595_by_group <- protein_long %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(
    n_wells  = n(),
    mean_595 = mean(abs_595),
    sd_595   = sd(abs_595),
    cv_595   = 100 * sd_595 / mean_595,
    .groups  = "drop"
  )

cat("--- sd_cv_595_by_group ---\n")
str(sd_cv_595_by_group)
--- sd_cv_595_by_group ---
tibble [24 × 8] (S3: tbl_df/tbl/data.frame)
 $ well_id       : chr [1:24] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name          : chr [1:24] "F07_01_36C" "F07_02_ambient" "F07_03_ambient" "F07_04_ambient" ...
 $ type          : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ std_conc_known: int [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells       : int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595      : num [1:24] 1.04 1.25 1.29 1.02 1.04 ...
 $ sd_595        : num [1:24] 0.0265 0.0505 0.073 0.0835 0.0481 ...
 $ cv_595        : num [1:24] 2.55 4.03 5.66 8.2 4.64 ...

6.2 SD and CV of concentration

sd_cv_conc_by_group <- protein_long %>%
  filter(!gen5_concentration_censored) %>%
  group_by(well_id, name, type, std_conc_known) %>%
  summarise(
    n_wells_valid_conc  = n(),
    mean_concentration  = mean(gen5_concentration_numeric),
    sd_concentration    = sd(gen5_concentration_numeric),
    cv_concentration    = 100 * sd_concentration / mean_concentration,
    .groups = "drop"
  )

cat("--- sd_cv_conc_by_group ---\n")
str(sd_cv_conc_by_group)
--- sd_cv_conc_by_group ---
tibble [23 × 8] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:23] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name              : chr [1:23] "F07_01_36C" "F07_02_ambient" "F07_03_ambient" "F07_04_ambient" ...
 $ type              : chr [1:23] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:23] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells_valid_conc: int [1:23] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_concentration: num [1:23] 694 1096 1168 652 691 ...
 $ sd_concentration  : num [1:23] 49.6 95.8 138.6 157.5 91.5 ...
 $ cv_concentration  : num [1:23] 7.14 8.74 11.87 24.15 13.24 ...

6.3 Combined per-group summary

group_stats <- sd_cv_595_by_group %>%
  left_join(
    sd_cv_conc_by_group %>%
      select(well_id, mean_concentration, sd_concentration, cv_concentration, n_wells_valid_conc),
    by = "well_id"
  ) %>%
  arrange(type, well_id)

cat("--- group_stats: one row per sample/standard, 595- and concentration-level stats ---\n")
str(group_stats)
kable(group_stats, digits = 3)
--- group_stats: one row per sample/standard, 595- and concentration-level stats ---
tibble [24 × 12] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:24] "SPL1" "SPL10" "SPL11" "SPL12" ...
 $ name              : chr [1:24] "F07_01_36C" "F07_02_ambient" "F07_03_ambient" "F07_04_ambient" ...
 $ type              : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells           : int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595          : num [1:24] 1.04 1.25 1.29 1.02 1.04 ...
 $ sd_595            : num [1:24] 0.0265 0.0505 0.073 0.0835 0.0481 ...
 $ cv_595            : num [1:24] 2.55 4.03 5.66 8.2 4.64 ...
 $ mean_concentration: num [1:24] 694 1096 1168 652 691 ...
 $ sd_concentration  : num [1:24] 49.6 95.8 138.6 157.5 91.5 ...
 $ cv_concentration  : num [1:24] 7.14 8.74 11.87 24.15 13.24 ...
 $ n_wells_valid_conc: int [1:24] 3 3 3 3 3 3 3 3 3 3 ...
well_id name type std_conc_known n_wells mean_595 sd_595 cv_595 mean_concentration sd_concentration cv_concentration n_wells_valid_conc
SPL1 F07_01_36C sample NA 3 1.040 0.027 2.549 694.467 49.577 7.139 3
SPL10 F07_02_ambient sample NA 3 1.252 0.051 4.035 1095.807 95.759 8.739 3
SPL11 F07_03_ambient sample NA 3 1.290 0.073 5.658 1168.307 138.620 11.865 3
SPL12 F07_04_ambient sample NA 3 1.018 0.083 8.203 652.091 157.509 24.154 3
SPL13 F07_05_ambient sample NA 3 1.038 0.048 4.637 690.741 91.456 13.240 3
SPL14 F07_06_ambient sample NA 3 1.296 0.099 7.647 1178.349 187.114 15.879 3
SPL15 F07_07_ambient sample NA 3 1.538 0.060 3.922 1638.927 114.400 6.980 3
SPL16 F07_08_ambient sample NA 3 0.988 0.022 2.184 595.253 41.154 6.914 3
SPL2 F07_02_36C sample NA 3 1.177 0.043 3.635 953.017 81.025 8.502 3
SPL3 F07_03_36C sample NA 3 1.315 0.024 1.841 1216.746 46.107 3.789 3
SPL4 F07_04_36C sample NA 3 1.499 0.016 1.048 1564.027 29.665 1.897 3
SPL5 F07_05_36C sample NA 3 0.878 0.031 3.585 388.299 59.694 15.373 3
SPL6 F07_06_36C sample NA 3 0.955 0.008 0.813 533.741 14.556 2.727 3
SPL7 F07_07_36C sample NA 3 1.230 0.033 2.689 1054.821 63.068 5.979 3
SPL8 F07_08_36C sample NA 3 0.850 0.023 2.675 334.997 42.754 12.762 3
SPL9 F07_01_ambient sample NA 3 0.805 0.059 7.285 249.551 110.645 44.338 3
STD1 BSA standard 2000 3 1.600 0.023 1.442 1755.192 43.330 2.469 3
STD2 BSA standard 1500 3 1.492 0.021 1.423 1549.754 40.397 2.607 3
STD3 BSA standard 1000 3 1.342 0.031 2.328 1266.321 59.723 4.716 3
STD4 BSA standard 750 3 1.164 0.039 3.324 930.092 73.726 7.927 3
STD5 BSA standard 500 3 0.970 0.024 2.455 561.213 44.720 7.968 3
STD6 BSA standard 250 3 0.796 0.030 3.726 232.689 55.661 23.921 3
STD7 BSA standard 125 3 0.686 0.016 2.314 23.525 29.610 125.869 3
STD8 BSA standard 0 3 0.571 0.007 1.165 NA NA NA NA

7 CV-BASED EXCLUSION

Groups with CV(concentration) > 15% – or an undefined concentration CV due to insufficient valid (non-censored) replicates – are flagged and excluded from all downstream analysis, including standard-curve fitting. The zero BSA standard (STD8) is exempt from this rule (see Background) and is always retained.

group_stats <- group_stats %>%
  mutate(
    is_zero_standard = type == "standard" & !is.na(std_conc_known) & std_conc_known == 0,
    high_cv = !is_zero_standard & (is.na(cv_concentration) | cv_concentration > 15)
  )

excluded_groups <- group_stats %>% filter(high_cv)

cat("--- excluded_groups: CV(concentration) > 15% (or undefined), dropped from further analysis ---\n")
str(excluded_groups)
kable(excluded_groups %>%
        select(well_id, name, type, std_conc_known, n_wells_valid_conc,
               mean_concentration, sd_concentration, cv_concentration),
      digits = 3)

clean_group_stats <- group_stats %>% filter(!high_cv)

cat("\nRetained after CV exclusion:", nrow(clean_group_stats), "of", nrow(group_stats), "groups\n")
str(clean_group_stats)
--- excluded_groups: CV(concentration) > 15% (or undefined), dropped from further analysis ---
tibble [6 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:6] "SPL12" "SPL14" "SPL5" "SPL9" ...
 $ name              : chr [1:6] "F07_04_ambient" "F07_06_ambient" "F07_05_36C" "F07_01_ambient" ...
 $ type              : chr [1:6] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:6] NA NA NA NA 250 125
 $ n_wells           : int [1:6] 3 3 3 3 3 3
 $ mean_595          : num [1:6] 1.018 1.296 0.878 0.805 0.796 ...
 $ sd_595            : num [1:6] 0.0835 0.0991 0.0315 0.0586 0.0297 ...
 $ cv_595            : num [1:6] 8.2 7.65 3.59 7.28 3.73 ...
 $ mean_concentration: num [1:6] 652 1178 388 250 233 ...
 $ sd_concentration  : num [1:6] 157.5 187.1 59.7 110.6 55.7 ...
 $ cv_concentration  : num [1:6] 24.2 15.9 15.4 44.3 23.9 ...
 $ n_wells_valid_conc: int [1:6] 3 3 3 3 3 3
 $ is_zero_standard  : logi [1:6] FALSE FALSE FALSE FALSE FALSE FALSE
 $ high_cv           : logi [1:6] TRUE TRUE TRUE TRUE TRUE TRUE
well_id name type std_conc_known n_wells_valid_conc mean_concentration sd_concentration cv_concentration
SPL12 F07_04_ambient sample NA 3 652.091 157.509 24.154
SPL14 F07_06_ambient sample NA 3 1178.349 187.114 15.879
SPL5 F07_05_36C sample NA 3 388.299 59.694 15.373
SPL9 F07_01_ambient sample NA 3 249.551 110.645 44.338
STD6 BSA standard 250 3 232.689 55.661 23.921
STD7 BSA standard 125 3 23.525 29.610 125.869 
Retained after CV exclusion: 18 of 24 groups
tibble [18 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:18] "SPL1" "SPL10" "SPL11" "SPL13" ...
 $ name              : chr [1:18] "F07_01_36C" "F07_02_ambient" "F07_03_ambient" "F07_05_ambient" ...
 $ type              : chr [1:18] "sample" "sample" "sample" "sample" ...
 $ std_conc_known    : int [1:18] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells           : int [1:18] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595          : num [1:18] 1.04 1.25 1.29 1.04 1.54 ...
 $ sd_595            : num [1:18] 0.0265 0.0505 0.073 0.0481 0.0603 ...
 $ cv_595            : num [1:18] 2.55 4.03 5.66 4.64 3.92 ...
 $ mean_concentration: num [1:18] 694 1096 1168 691 1639 ...
 $ sd_concentration  : num [1:18] 49.6 95.8 138.6 91.5 114.4 ...
 $ cv_concentration  : num [1:18] 7.14 8.74 11.87 13.24 6.98 ...
 $ n_wells_valid_conc: int [1:18] 3 3 3 3 3 3 3 3 3 3 ...
 $ is_zero_standard  : logi [1:18] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv           : logi [1:18] FALSE FALSE FALSE FALSE FALSE FALSE ...

8 STANDARD CURVE

Fit using only the retained (CV ≤ 15%) standard points: known BSA concentration (Conc/Dil) on the x-axis, mean 595 nm absorbance on the y-axis.

standards_clean <- clean_group_stats %>% filter(type == "standard")

cat("--- standards_clean: standard points retained for curve fitting ---\n")
str(standards_clean)

standard_curve_model <- lm(mean_595 ~ std_conc_known, data = standards_clean)

curve_slope     <- unname(coef(standard_curve_model)[2])
curve_intercept <- unname(coef(standard_curve_model)[1])
curve_r2        <- summary(standard_curve_model)$r.squared

cat("Standard curve fit: 595 =", format(curve_slope, digits = 6), "* concentration +",
    format(curve_intercept, digits = 4), "\n")
cat("R2:", round(curve_r2, 5), "\n")
--- standards_clean: standard points retained for curve fitting ---
tibble [6 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:6] "STD1" "STD2" "STD3" "STD4" ...
 $ name              : chr [1:6] "BSA" "BSA" "BSA" "BSA" ...
 $ type              : chr [1:6] "standard" "standard" "standard" "standard" ...
 $ std_conc_known    : int [1:6] 2000 1500 1000 750 500 0
 $ n_wells           : int [1:6] 3 3 3 3 3 3
 $ mean_595          : num [1:6] 1.6 1.49 1.34 1.16 0.97 ...
 $ sd_595            : num [1:6] 0.0231 0.0212 0.0312 0.0387 0.0238 ...
 $ cv_595            : num [1:6] 1.44 1.42 2.33 3.32 2.45 ...
 $ mean_concentration: num [1:6] 1755 1550 1266 930 561 ...
 $ sd_concentration  : num [1:6] 43.3 40.4 59.7 73.7 44.7 ...
 $ cv_concentration  : num [1:6] 2.47 2.61 4.72 7.93 7.97 ...
 $ n_wells_valid_conc: int [1:6] 3 3 3 3 3 NA
 $ is_zero_standard  : logi [1:6] FALSE FALSE FALSE FALSE FALSE TRUE
 $ high_cv           : logi [1:6] FALSE FALSE FALSE FALSE FALSE FALSE
Standard curve fit: 595 = 0.000507804 * concentration + 0.7032 
R2: 0.92166

8.1 Plot standard curve

eq_label <- paste0("y = ", format(curve_slope, digits = 4, scientific = TRUE), "x + ",
                    sprintf("%.4f", curve_intercept),
                    "\nR² = ", sprintf("%.4f", curve_r2))

standard_curve_plot <- ggplot(standards_clean, aes(x = std_conc_known, y = mean_595)) +
  geom_smooth(method = "lm", formula = y ~ x, se = FALSE, color = "steelblue", linewidth = 0.6) +
  geom_errorbar(aes(ymin = mean_595 - sd_595, ymax = mean_595 + sd_595),
                width = 30, color = "grey40") +
  geom_point(size = 2.6, color = "steelblue") +
  annotate("text", x = -Inf, y = Inf, label = eq_label,
           hjust = -0.08, vjust = 1.3, size = 3.6, fontface = "bold") +
  labs(x = "BSA standard concentration (µg/mL)",
       y = "Mean absorbance (595 nm)",
       title = "BSA standard curve",
       subtitle = "Points with technical-replicate CV(concentration) > 15% excluded (see below)") +
  theme_bw(base_size = 12)

print(standard_curve_plot)

ggsave(file.path(output_dir, "standard_curve_595.png"), standard_curve_plot,
       width = 7, height = 5.5, dpi = 300)

Standard points excluded from the curve:

excluded_standards <- group_stats %>% filter(type == "standard", high_cv)

cat("--- excluded_standards ---\n")
str(excluded_standards)
kable(excluded_standards %>%
        select(well_id, std_conc_known, mean_concentration, sd_concentration, cv_concentration),
      digits = 3)
--- excluded_standards ---
tibble [2 × 14] (S3: tbl_df/tbl/data.frame)
 $ well_id           : chr [1:2] "STD6" "STD7"
 $ name              : chr [1:2] "BSA" "BSA"
 $ type              : chr [1:2] "standard" "standard"
 $ std_conc_known    : int [1:2] 250 125
 $ n_wells           : int [1:2] 3 3
 $ mean_595          : num [1:2] 0.796 0.686
 $ sd_595            : num [1:2] 0.0297 0.0159
 $ cv_595            : num [1:2] 3.73 2.31
 $ mean_concentration: num [1:2] 232.7 23.5
 $ sd_concentration  : num [1:2] 55.7 29.6
 $ cv_concentration  : num [1:2] 23.9 125.9
 $ n_wells_valid_conc: int [1:2] 3 3
 $ is_zero_standard  : logi [1:2] FALSE FALSE
 $ high_cv           : logi [1:2] TRUE TRUE
well_id std_conc_known mean_concentration sd_concentration cv_concentration
STD6 250 232.689 55.661 23.921
STD7 125 23.525 29.610 125.869

9 APPLYING THE STANDARD CURVE TO SAMPLES

Samples that passed CV QC are back-calculated against the fitted curve. Samples whose mean 595 falls outside the retained standards’ absorbance range are flagged as extrapolated.

samples_clean <- clean_group_stats %>%
  filter(type == "sample") %>%
  mutate(
    se_595                    = sd_595 / sqrt(n_wells),
    calculated_concentration  = (mean_595 - curve_intercept) / curve_slope,
    in_curve_range            = mean_595 >= min(standards_clean$mean_595) &
                                  mean_595 <= max(standards_clean$mean_595)
  ) %>%
  arrange(name)

cat("--- samples_clean: CV-QC-passing samples, quantified against the fitted standard curve ---\n")
str(samples_clean)
--- samples_clean: CV-QC-passing samples, quantified against the fitted standard curve ---
tibble [12 × 17] (S3: tbl_df/tbl/data.frame)
 $ well_id                 : chr [1:12] "SPL1" "SPL2" "SPL10" "SPL3" ...
 $ name                    : chr [1:12] "F07_01_36C" "F07_02_36C" "F07_02_ambient" "F07_03_36C" ...
 $ type                    : chr [1:12] "sample" "sample" "sample" "sample" ...
 $ std_conc_known          : int [1:12] NA NA NA NA NA NA NA NA NA NA ...
 $ n_wells                 : int [1:12] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595                : num [1:12] 1.04 1.18 1.25 1.32 1.29 ...
 $ sd_595                  : num [1:12] 0.0265 0.0428 0.0505 0.0242 0.073 ...
 $ cv_595                  : num [1:12] 2.55 3.63 4.03 1.84 5.66 ...
 $ mean_concentration      : num [1:12] 694 953 1096 1217 1168 ...
 $ sd_concentration        : num [1:12] 49.6 81 95.8 46.1 138.6 ...
 $ cv_concentration        : num [1:12] 7.14 8.5 8.74 3.79 11.87 ...
 $ n_wells_valid_conc      : int [1:12] 3 3 3 3 3 3 3 3 3 3 ...
 $ is_zero_standard        : logi [1:12] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv                 : logi [1:12] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ se_595                  : num [1:12] 0.0153 0.0247 0.0292 0.014 0.0421 ...
 $ calculated_concentration: num [1:12] 663 932 1080 1205 1156 ...
 $ in_curve_range          : logi [1:12] TRUE TRUE TRUE TRUE TRUE TRUE ...
cat("Samples with absorbance outside the retained standard curve's range (extrapolated):\n")
kable(samples_clean %>% filter(!in_curve_range) %>%
        select(well_id, name, mean_595, calculated_concentration),
      digits = 3)
Samples with absorbance outside the retained standard curve's range (extrapolated):

well_id name mean_595 calculated_concentration ——— —— ———- ————————–

9.1 Plot samples on standard curve

Every QC-passing sample is plotted at its own mean 595 and back-calculated concentration, with vertical error bars of ±1 standard error (of mean 595, across its 2-3 retained technical replicates), overlaid on the standard curve fit line and retained standard points.

samples_on_curve_plot <- ggplot() +
  geom_smooth(data = standards_clean, aes(x = std_conc_known, y = mean_595),
              method = "lm", formula = y ~ x, se = FALSE,
              color = "steelblue", linewidth = 0.6) +
  geom_point(data = standards_clean, aes(x = std_conc_known, y = mean_595),
             size = 2.6, color = "steelblue") +
  geom_errorbar(data = samples_clean,
                aes(x = calculated_concentration,
                    ymin = mean_595 - se_595, ymax = mean_595 + se_595),
                width = 30, color = "grey40") +
  geom_point(data = samples_clean,
             aes(x = calculated_concentration, y = mean_595, shape = in_curve_range),
             size = 2.4, color = "firebrick", alpha = 0.85) +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                      labels = c(`TRUE` = "in range", `FALSE` = "extrapolated"),
                      name = "Sample status") +
  labs(x = "Concentration (µg/mL)",
       y = "Mean absorbance (595 nm)",
       title = "BSA standard curve with QC-passing samples overlaid",
       subtitle = "Samples plotted at their back-calculated concentration; error bars = ±1 SE of mean 595") +
  theme_bw(base_size = 12)

print(samples_on_curve_plot)

ggsave(file.path(output_dir, "standard_curve_with_samples.png"), samples_on_curve_plot,
       width = 8, height = 5.5, dpi = 300)

10 RESULTS TABLE

results_table <- samples_clean %>%
  transmute(
    Sample                              = name,
    `Well ID`                           = well_id,
    `Mean 595`                          = round(mean_595, 3),
    `CV 595 (%)`                        = round(cv_595, 2),
    `CV concentration (%)`              = round(cv_concentration, 2),
    `Calculated concentration (ug/mL)`  = round(calculated_concentration, 1),
    `In standard curve range`           = ifelse(in_curve_range, "yes", "NO - extrapolated")
  ) %>%
  arrange(gsub("[0-9]+$", "", `Well ID`), as.numeric(gsub("^[A-Za-z]+", "", `Well ID`)))

kable(results_table)
Sample Well ID Mean 595 CV 595 (%) CV concentration (%) Calculated concentration (ug/mL) In standard curve range
F07_01_36C SPL1 1.040 2.55 7.14 663.3 yes
F07_02_36C SPL2 1.177 3.63 8.50 932.4 yes
F07_03_36C SPL3 1.315 1.84 3.79 1205.5 yes
F07_04_36C SPL4 1.499 1.05 1.90 1567.2 yes
F07_06_36C SPL6 0.955 0.81 2.73 496.5 yes
F07_07_36C SPL7 1.230 2.69 5.98 1038.1 yes
F07_08_36C SPL8 0.850 2.67 12.76 289.8 yes
F07_02_ambient SPL10 1.252 4.03 8.74 1080.1 yes
F07_03_ambient SPL11 1.290 5.66 11.87 1155.6 yes
F07_05_ambient SPL13 1.038 4.64 13.24 659.3 yes
F07_07_ambient SPL15 1.538 3.92 6.98 1644.6 yes
F07_08_ambient SPL16 0.988 2.18 6.91 560.2 yes

Only samples passing all QC – CV ≤ 15% and within the retained standard curve’s range (i.e. not extrapolated) – are written to the CSV output.

results_table_export <- results_table %>% filter(`In standard curve range` == "yes")

cat("--- results_table_export: samples passing all QC (CV and in-range), written to CSV ---\n")
str(results_table_export)

write.csv(results_table_export, file.path(output_dir, "sample_protein_concentrations.csv"), row.names = FALSE)
--- results_table_export: samples passing all QC (CV and in-range), written to CSV ---
tibble [12 × 7] (S3: tbl_df/tbl/data.frame)
 $ Sample                          : chr [1:12] "F07_01_36C" "F07_02_36C" "F07_03_36C" "F07_04_36C" ...
 $ Well ID                         : chr [1:12] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ Mean 595                        : num [1:12] 1.04 1.177 1.315 1.499 0.955 ...
 $ CV 595 (%)                      : num [1:12] 2.55 3.63 1.84 1.05 0.81 2.69 2.67 4.03 5.66 4.64 ...
 $ CV concentration (%)            : num [1:12] 7.14 8.5 3.79 1.9 2.73 ...
 $ Calculated concentration (ug/mL): num [1:12] 663 932 1206 1567 496 ...
 $ In standard curve range         : chr [1:12] "yes" "yes" "yes" "yes" ...

11 SUMMARY

  • 24 sample/standard triplicate groups were imported (8 BSA standards, 16 samples).
  • 6 group(s) exceeded 15% technical-replicate CV(concentration), or had an undefined CV from insufficient valid replicates, and were excluded from all downstream analysis: SPL12 (F07_04_ambient), SPL14 (F07_06_ambient), SPL5 (F07_05_36C), SPL9 (F07_01_ambient), STD6 (BSA), STD7 (BSA).
  • The BSA standard curve was fit on the 6 remaining standard points: slope = 0.0005078, intercept = 0.7032, R² = 0.9217.
  • 0 CV-QC-passing sample(s) fell outside the retained standard curve’s absorbance range and are flagged as extrapolated in the results table above.

11.1 Samples passing QC

Sample names are sorted naturally (e.g. F07_02_ambient before F07_10_ambient, and ignoring any stray punctuation), rather than by plain lexicographic string order.

natural_sort_key <- function(x) gsub("[^A-Za-z0-9]", "", x)
passing_samples_table <- samples_clean %>%
  transmute(sample_name = name,
            `concentration(ug/mL)` = round(calculated_concentration, 1)) %>%
  arrange(natural_sort_key(sample_name))

cat("--- passing_samples_table: samples passing CV QC, name + calculated concentration ---\n")
str(passing_samples_table)

kable(passing_samples_table)
--- passing_samples_table: samples passing CV QC, name + calculated concentration ---
tibble [12 × 2] (S3: tbl_df/tbl/data.frame)
 $ sample_name         : chr [1:12] "F07_01_36C" "F07_02_36C" "F07_02_ambient" "F07_03_36C" ...
 $ concentration(ug/mL): num [1:12] 663 932 1080 1206 1156 ...
sample_name concentration(ug/mL)
F07_01_36C 663.3
F07_02_36C 932.4
F07_02_ambient 1080.1
F07_03_36C 1205.5
F07_03_ambient 1155.6
F07_04_36C 1567.2
F07_05_ambient 659.3
F07_06_36C 496.5
F07_07_36C 1038.1
F07_07_ambient 1644.6
F07_08_36C 289.8
F07_08_ambient 560.2

SAMPLE REASSAYS

1 BACKGROUND

Protein quantification (BCA/Bradford-style, 595 nm) of Magallana gigas (Pacific oyster) SoRMI re-assay plate, 9 samples plus an 8-point BSA standard curve, each measured in triplicate wells. All 9 samples are re-assays of samples previously flagged for re-assay in the F05 and F07 analyses (5 from F05 – CV or standard-curve-range failures – and 4 from F07 – CV failures), now run together on a single plate against a fresh standard curve.

The raw Gen5 export (Gen5-20260813-mgig-BSA-F05-F07-reassay-absorbance.csv) has the same per-well layout as the earlier F05/F07 exports (see those documents for detail), but with one new wrinkle: sample names carry a dilution-factor suffix, <sample>-df.<N>, e.g. F05_01_ambient-df.0 or F05_03_ambient-df.2. df.0 means the well was read on undiluted material; df.N (N > 0) means the material was diluted N-fold before reading, so the well’s back-calculated concentration must be multiplied by N to recover the sample’s true (pre-dilution) concentration – df.0 is therefore treated as a multiplier of 1, not 0. This multiplication is kept as a distinct, explicit step (see “Applying the standard curve to samples” below) so the well-level (as-measured) and final (dilution-corrected) concentrations both remain visible.

As before, this document ignores the instrument’s pre-computed summary columns and recomputes mean, SD, and CV independently in R via dplyr, for both the raw absorbance (595 nm) and the concentration.

CV QC metric used for exclusion: the >15% technical-replicate CV exclusion rule (including for standard-curve points) is applied to CV of concentration (matching what the instrument’s own CV (%) column already represents for this export – verified against the raw file: e.g. SPL1’s given Mean/SD/CV of 883.966/50.075/5.665 are exactly the mean/SD/CV of its three [Concentration] values, not of its 595 values, and this holds regardless of that sample’s dilution factor since the raw [Concentration] column reflects the as-measured, pre-correction well concentration).

Two edge cases in the raw [Concentration] column need explicit handling:

  • Censored reads. Wells reading outside the instrument’s own internal standard-curve range are reported as <0.000 or >2100.000 rather than a number. These are parsed to a _censored flag plus the boundary value, and excluded from concentration mean/SD/CV calculations (their true value is unknown, not equal to the boundary).
  • The zero (blank) BSA standard (STD8) is always censored (<0.000 on all three wells), by construction – a true-zero standard’s back-calculated concentration is not a meaningful quantity to compute a CV on, and this is not a data-quality problem. STD8 is therefore exempt from the concentration-CV exclusion rule and is retained as the curve’s origin point on the strength of its (excellent) 595 CV instead.

Any other group left with fewer than 2 valid (non-censored) concentration replicates has an undefined (NA) concentration CV and is treated as failing QC, consistent with the source file itself being unable to compute a CV there either (?????). In this plate’s data, every sample well yields a numeric (non-censored) concentration, so this case only arises for the zero BSA standard (which is exempted, above).

2 SETUP

2.1 Libraries

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Set output directory

output_dir <- "../outputs/Gen5-20260813-mgig-BSA-F05-F07-reassay"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

3 DATA IMPORT

The raw file is a Gen5 per-well export with non-syntactic column names ([Concentration], CV (%), etc.) and a UTF-8 byte-order mark on the first header cell, so columns are renamed by position immediately after import rather than referenced by their original names.

data_path <- "../data/BSA/raw_absorbance/Gen5-20260813-mgig-BSA-F05-F07-reassay-absorbance.csv"

protein_raw <- read.csv(data_path, stringsAsFactors = FALSE, check.names = FALSE,
                         na.strings = c("", "NA"))
colnames(protein_raw) <- c("well_id", "name", "well", "std_conc_known", "abs_595",
                           "gen5_concentration", "gen5_count", "gen5_mean",
                           "gen5_sd", "gen5_cv")

cat("--- protein_raw: as imported from the Gen5 export ---\n")
str(protein_raw)
--- protein_raw: as imported from the Gen5 export ---
'data.frame':   51 obs. of  10 variables:
 $ well_id           : chr  "SPL1" NA NA "SPL2" ...
 $ name              : chr  "F05_01_ambient-df.0" NA NA "F05_02_ambient-df.0" ...
 $ well              : chr  "B1" "B2" "B3" "B4" ...
 $ std_conc_known    : int  NA NA NA NA NA NA NA NA NA NA ...
 $ abs_595           : num  1.14 1.18 1.19 1.22 1.28 ...
 $ gen5_concentration: chr  "827.572" "901.105" "923.221" "989.193" ...
 $ gen5_count        : int  3 NA NA 3 NA NA 3 NA NA 3 ...
 $ gen5_mean         : chr  "883.966" NA NA "1044.327" ...
 $ gen5_sd           : chr  "50.075" NA NA "56.98" ...
 $ gen5_cv           : chr  "5.665" NA NA "5.456" ...

4 RESHAPING

well_id and name are only populated on the first row of each triplicate group in the raw export, so they are filled downward. Each row is also classified as a standard (well_id starts with STD) or a sample (SPL), and the instrument’s censored concentration strings (<0.000, >2100.000) are parsed into a numeric value plus a gen5_concentration_censored flag. For sample rows, the -df.<N> suffix is parsed off name into a clean base_sample_name, plus dilution_code (the raw N) and dilution_multiplier (1 when N == 0, otherwise N).

protein_long <- protein_raw %>%
  select(well_id, name, well, std_conc_known, abs_595, gen5_concentration) %>%
  fill(well_id, name, .direction = "down") %>%
  mutate(
    type = if_else(grepl("^STD", well_id), "standard", "sample"),
    gen5_concentration_censored = grepl("^[<>]", gen5_concentration),
    gen5_concentration_numeric  = as.numeric(gsub("[<>]", "", gen5_concentration)),
    base_sample_name = if_else(type == "sample",
                                sub("-df\\.[0-9]+$", "", name),
                                NA_character_),
    dilution_code = if_else(type == "sample",
                             as.numeric(sub("^.*-df\\.([0-9]+)$", "\\1", name)),
                             NA_real_),
    dilution_multiplier = if_else(type == "sample",
                                   if_else(dilution_code == 0, 1, dilution_code),
                                   NA_real_)
  )

cat("--- protein_long: one row per well, group identifiers filled down, dilution parsed ---\n")
str(protein_long)
--- protein_long: one row per well, group identifiers filled down, dilution parsed ---
'data.frame':   51 obs. of  12 variables:
 $ well_id                    : chr  "SPL1" "SPL1" "SPL1" "SPL2" ...
 $ name                       : chr  "F05_01_ambient-df.0" "F05_01_ambient-df.0" "F05_01_ambient-df.0" "F05_02_ambient-df.0" ...
 $ well                       : chr  "B1" "B2" "B3" "B4" ...
 $ std_conc_known             : int  NA NA NA NA NA NA NA NA NA NA ...
 $ abs_595                    : num  1.14 1.18 1.19 1.22 1.28 ...
 $ gen5_concentration         : chr  "827.572" "901.105" "923.221" "989.193" ...
 $ type                       : chr  "sample" "sample" "sample" "sample" ...
 $ gen5_concentration_censored: logi  FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ gen5_concentration_numeric : num  828 901 923 989 1103 ...
 $ base_sample_name           : chr  "F05_01_ambient" "F05_01_ambient" "F05_01_ambient" "F05_02_ambient" ...
 $ dilution_code              : num  0 0 0 0 0 0 2 2 2 2 ...
 $ dilution_multiplier        : num  1 1 1 1 1 1 2 2 2 2 ...

5 MEAN VALUES

5.1 Mean 595 nm absorbance

mean_595_by_group <- protein_long %>%
  group_by(well_id, name, type, std_conc_known,
           base_sample_name, dilution_code, dilution_multiplier) %>%
  summarise(n_wells  = n(),
            mean_595 = mean(abs_595),
            .groups  = "drop")

cat("--- mean_595_by_group ---\n")
str(mean_595_by_group)
--- mean_595_by_group ---
tibble [17 × 9] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:17] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name               : chr [1:17] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type               : chr [1:17] "sample" "sample" "sample" "sample" ...
 $ std_conc_known     : int [1:17] NA NA NA NA NA NA NA NA NA 2000 ...
 $ base_sample_name   : chr [1:17] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code      : num [1:17] 0 0 2 2 2 0 0 0 0 NA ...
 $ dilution_multiplier: num [1:17] 1 1 2 2 2 1 1 1 1 NA ...
 $ n_wells            : int [1:17] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595           : num [1:17] 1.17 1.25 1.2 1.29 1.41 ...

5.2 Mean concentration

Censored wells (reading outside the instrument’s own standard-curve range) are dropped before averaging, since their true concentration is unknown.

mean_conc_by_group <- protein_long %>%
  filter(!gen5_concentration_censored) %>%
  group_by(well_id, name, type, std_conc_known,
           base_sample_name, dilution_code, dilution_multiplier) %>%
  summarise(n_wells_valid_conc  = n(),
            mean_concentration  = mean(gen5_concentration_numeric),
            .groups = "drop")

cat("--- mean_conc_by_group ---\n")
str(mean_conc_by_group)
--- mean_conc_by_group ---
tibble [16 × 9] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:16] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name               : chr [1:16] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type               : chr [1:16] "sample" "sample" "sample" "sample" ...
 $ std_conc_known     : int [1:16] NA NA NA NA NA NA NA NA NA 2000 ...
 $ base_sample_name   : chr [1:16] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code      : num [1:16] 0 0 2 2 2 0 0 0 0 NA ...
 $ dilution_multiplier: num [1:16] 1 1 2 2 2 1 1 1 1 NA ...
 $ n_wells_valid_conc : int [1:16] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_concentration : num [1:16] 884 1044 956 1120 1353 ...

6 STANDARD DEVIATION AND COEFFICIENT OF VARIATION

6.1 SD and CV of 595 nm absorbance

sd_cv_595_by_group <- protein_long %>%
  group_by(well_id, name, type, std_conc_known,
           base_sample_name, dilution_code, dilution_multiplier) %>%
  summarise(
    n_wells  = n(),
    mean_595 = mean(abs_595),
    sd_595   = sd(abs_595),
    cv_595   = 100 * sd_595 / mean_595,
    .groups  = "drop"
  )

cat("--- sd_cv_595_by_group ---\n")
str(sd_cv_595_by_group)
--- sd_cv_595_by_group ---
tibble [17 × 11] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:17] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name               : chr [1:17] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type               : chr [1:17] "sample" "sample" "sample" "sample" ...
 $ std_conc_known     : int [1:17] NA NA NA NA NA NA NA NA NA 2000 ...
 $ base_sample_name   : chr [1:17] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code      : num [1:17] 0 0 2 2 2 0 0 0 0 NA ...
 $ dilution_multiplier: num [1:17] 1 1 2 2 2 1 1 1 1 NA ...
 $ n_wells            : int [1:17] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595           : num [1:17] 1.17 1.25 1.2 1.29 1.41 ...
 $ sd_595             : num [1:17] 0.0267 0.03 0.0259 0.0375 0.0229 ...
 $ cv_595             : num [1:17] 2.29 2.4 2.15 2.91 1.62 ...

6.2 SD and CV of concentration

sd_cv_conc_by_group <- protein_long %>%
  filter(!gen5_concentration_censored) %>%
  group_by(well_id, name, type, std_conc_known,
           base_sample_name, dilution_code, dilution_multiplier) %>%
  summarise(
    n_wells_valid_conc  = n(),
    mean_concentration  = mean(gen5_concentration_numeric),
    sd_concentration    = sd(gen5_concentration_numeric),
    cv_concentration    = 100 * sd_concentration / mean_concentration,
    .groups = "drop"
  )

cat("--- sd_cv_conc_by_group ---\n")
str(sd_cv_conc_by_group)
--- sd_cv_conc_by_group ---
tibble [16 × 11] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:16] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name               : chr [1:16] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type               : chr [1:16] "sample" "sample" "sample" "sample" ...
 $ std_conc_known     : int [1:16] NA NA NA NA NA NA NA NA NA 2000 ...
 $ base_sample_name   : chr [1:16] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code      : num [1:16] 0 0 2 2 2 0 0 0 0 NA ...
 $ dilution_multiplier: num [1:16] 1 1 2 2 2 1 1 1 1 NA ...
 $ n_wells_valid_conc : int [1:16] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_concentration : num [1:16] 884 1044 956 1120 1353 ...
 $ sd_concentration   : num [1:16] 50.1 57 48.4 71.7 43.2 ...
 $ cv_concentration   : num [1:16] 5.66 5.46 5.06 6.4 3.19 ...

6.3 Combined per-group summary

group_stats <- sd_cv_595_by_group %>%
  left_join(
    sd_cv_conc_by_group %>%
      select(well_id, mean_concentration, sd_concentration, cv_concentration, n_wells_valid_conc),
    by = "well_id"
  ) %>%
  arrange(type, well_id)

cat("--- group_stats: one row per sample/standard, 595- and concentration-level stats ---\n")
str(group_stats)
kable(group_stats, digits = 3)
--- group_stats: one row per sample/standard, 595- and concentration-level stats ---
tibble [17 × 15] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:17] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name               : chr [1:17] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type               : chr [1:17] "sample" "sample" "sample" "sample" ...
 $ std_conc_known     : int [1:17] NA NA NA NA NA NA NA NA NA 2000 ...
 $ base_sample_name   : chr [1:17] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code      : num [1:17] 0 0 2 2 2 0 0 0 0 NA ...
 $ dilution_multiplier: num [1:17] 1 1 2 2 2 1 1 1 1 NA ...
 $ n_wells            : int [1:17] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595           : num [1:17] 1.17 1.25 1.2 1.29 1.41 ...
 $ sd_595             : num [1:17] 0.0267 0.03 0.0259 0.0375 0.0229 ...
 $ cv_595             : num [1:17] 2.29 2.4 2.15 2.91 1.62 ...
 $ mean_concentration : num [1:17] 884 1044 956 1120 1353 ...
 $ sd_concentration   : num [1:17] 50.1 57 48.4 71.7 43.2 ...
 $ cv_concentration   : num [1:17] 5.66 5.46 5.06 6.4 3.19 ...
 $ n_wells_valid_conc : int [1:17] 3 3 3 3 3 3 3 3 3 3 ...
well_id name type std_conc_known base_sample_name dilution_code dilution_multiplier n_wells mean_595 sd_595 cv_595 mean_concentration sd_concentration cv_concentration n_wells_valid_conc
SPL1 F05_01_ambient-df.0 sample NA F05_01_ambient 0 1 3 1.166 0.027 2.287 883.966 50.075 5.665 3
SPL2 F05_02_ambient-df.0 sample NA F05_02_ambient 0 1 3 1.251 0.030 2.402 1044.327 56.980 5.456 3
SPL3 F05_03_ambient-df.2 sample NA F05_03_ambient 2 2 3 1.205 0.026 2.149 956.302 48.370 5.058 3
SPL4 F05_06_ambient-df.2 sample NA F05_06_ambient 2 2 3 1.291 0.038 2.907 1120.380 71.728 6.402 3
SPL5 F05_08_ambient-df.2 sample NA F05_08_ambient 2 2 3 1.414 0.023 1.619 1353.329 43.195 3.192 3
SPL6 F07_05_36C-df.0 sample NA F07_05_36C 0 1 3 0.932 0.008 0.810 441.760 13.784 3.120 3
SPL7 F07_01_ambient-df.0 sample NA F07_01_ambient 0 1 3 0.846 0.019 2.280 278.438 35.812 12.862 3
SPL8 F07_04_ambient-df.0 sample NA F07_04_ambient 0 1 3 1.001 0.026 2.551 571.561 48.935 8.562 3
SPL9 F07_06_ambient-df.0 sample NA F07_06_ambient 0 1 3 1.293 0.020 1.545 1123.972 37.008 3.293 3
STD1 BSA standard 2000 NA NA NA 3 1.620 0.017 1.059 1742.543 32.899 1.888 3
STD2 BSA standard 1500 NA NA NA 3 1.543 0.026 1.654 1596.044 48.516 3.040 3
STD3 BSA standard 1000 NA NA NA 3 1.374 0.011 0.772 1276.142 19.343 1.516 3
STD4 BSA standard 750 NA NA NA 3 1.146 0.029 2.553 845.908 55.390 6.548 3
STD5 BSA standard 500 NA NA NA 3 1.004 0.024 2.342 577.674 44.343 7.676 3
STD6 BSA standard 250 NA NA NA 3 0.811 0.009 1.054 212.656 16.562 7.788 3
STD7 BSA standard 125 NA NA NA 3 0.736 0.031 4.209 71.197 58.999 82.867 3
STD8 BSA standard 0 NA NA NA 3 0.594 0.014 2.297 NA NA NA NA

7 CV-BASED EXCLUSION

Groups with CV(concentration) > 15% – or an undefined concentration CV due to insufficient valid (non-censored) replicates – are flagged and excluded from all downstream analysis, including standard-curve fitting. The zero BSA standard (STD8) is exempt from this rule (see Background) and is always retained.

group_stats <- group_stats %>%
  mutate(
    is_zero_standard = type == "standard" & !is.na(std_conc_known) & std_conc_known == 0,
    high_cv = !is_zero_standard & (is.na(cv_concentration) | cv_concentration > 15)
  )

excluded_groups <- group_stats %>% filter(high_cv)

cat("--- excluded_groups: CV(concentration) > 15% (or undefined), dropped from further analysis ---\n")
str(excluded_groups)
kable(excluded_groups %>%
        select(well_id, name, type, std_conc_known, n_wells_valid_conc,
               mean_concentration, sd_concentration, cv_concentration),
      digits = 3)

clean_group_stats <- group_stats %>% filter(!high_cv)

cat("\nRetained after CV exclusion:", nrow(clean_group_stats), "of", nrow(group_stats), "groups\n")
str(clean_group_stats)
--- excluded_groups: CV(concentration) > 15% (or undefined), dropped from further analysis ---
tibble [1 × 17] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr "STD7"
 $ name               : chr "BSA"
 $ type               : chr "standard"
 $ std_conc_known     : int 125
 $ base_sample_name   : chr NA
 $ dilution_code      : num NA
 $ dilution_multiplier: num NA
 $ n_wells            : int 3
 $ mean_595           : num 0.736
 $ sd_595             : num 0.031
 $ cv_595             : num 4.21
 $ mean_concentration : num 71.2
 $ sd_concentration   : num 59
 $ cv_concentration   : num 82.9
 $ n_wells_valid_conc : int 3
 $ is_zero_standard   : logi FALSE
 $ high_cv            : logi TRUE
well_id name type std_conc_known n_wells_valid_conc mean_concentration sd_concentration cv_concentration
STD7 BSA standard 125 3 71.197 58.999 82.867 
Retained after CV exclusion: 16 of 17 groups
tibble [16 × 17] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:16] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name               : chr [1:16] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type               : chr [1:16] "sample" "sample" "sample" "sample" ...
 $ std_conc_known     : int [1:16] NA NA NA NA NA NA NA NA NA 2000 ...
 $ base_sample_name   : chr [1:16] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code      : num [1:16] 0 0 2 2 2 0 0 0 0 NA ...
 $ dilution_multiplier: num [1:16] 1 1 2 2 2 1 1 1 1 NA ...
 $ n_wells            : int [1:16] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_595           : num [1:16] 1.17 1.25 1.2 1.29 1.41 ...
 $ sd_595             : num [1:16] 0.0267 0.03 0.0259 0.0375 0.0229 ...
 $ cv_595             : num [1:16] 2.29 2.4 2.15 2.91 1.62 ...
 $ mean_concentration : num [1:16] 884 1044 956 1120 1353 ...
 $ sd_concentration   : num [1:16] 50.1 57 48.4 71.7 43.2 ...
 $ cv_concentration   : num [1:16] 5.66 5.46 5.06 6.4 3.19 ...
 $ n_wells_valid_conc : int [1:16] 3 3 3 3 3 3 3 3 3 3 ...
 $ is_zero_standard   : logi [1:16] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv            : logi [1:16] FALSE FALSE FALSE FALSE FALSE FALSE ...

8 STANDARD CURVE

Fit using only the retained (CV ≤ 15%) standard points: known BSA concentration (Conc/Dil) on the x-axis, mean 595 nm absorbance on the y-axis. Dilution correction (below) has no bearing on the curve fit itself, since standards are not diluted.

standards_clean <- clean_group_stats %>% filter(type == "standard")

cat("--- standards_clean: standard points retained for curve fitting ---\n")
str(standards_clean)

standard_curve_model <- lm(mean_595 ~ std_conc_known, data = standards_clean)

curve_slope     <- unname(coef(standard_curve_model)[2])
curve_intercept <- unname(coef(standard_curve_model)[1])
curve_r2        <- summary(standard_curve_model)$r.squared

cat("Standard curve fit: 595 =", format(curve_slope, digits = 6), "* concentration +",
    format(curve_intercept, digits = 4), "\n")
cat("R2:", round(curve_r2, 5), "\n")
--- standards_clean: standard points retained for curve fitting ---
tibble [7 × 17] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr [1:7] "STD1" "STD2" "STD3" "STD4" ...
 $ name               : chr [1:7] "BSA" "BSA" "BSA" "BSA" ...
 $ type               : chr [1:7] "standard" "standard" "standard" "standard" ...
 $ std_conc_known     : int [1:7] 2000 1500 1000 750 500 250 0
 $ base_sample_name   : chr [1:7] NA NA NA NA ...
 $ dilution_code      : num [1:7] NA NA NA NA NA NA NA
 $ dilution_multiplier: num [1:7] NA NA NA NA NA NA NA
 $ n_wells            : int [1:7] 3 3 3 3 3 3 3
 $ mean_595           : num [1:7] 1.62 1.54 1.37 1.15 1 ...
 $ sd_595             : num [1:7] 0.0172 0.0255 0.0106 0.0293 0.0235 ...
 $ cv_595             : num [1:7] 1.059 1.654 0.772 2.553 2.342 ...
 $ mean_concentration : num [1:7] 1743 1596 1276 846 578 ...
 $ sd_concentration   : num [1:7] 32.9 48.5 19.3 55.4 44.3 ...
 $ cv_concentration   : num [1:7] 1.89 3.04 1.52 6.55 7.68 ...
 $ n_wells_valid_conc : int [1:7] 3 3 3 3 3 3 NA
 $ is_zero_standard   : logi [1:7] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv            : logi [1:7] FALSE FALSE FALSE FALSE FALSE FALSE ...
Standard curve fit: 595 = 0.000522028 * concentration + 0.7086 
R2: 0.93363

8.1 Plot standard curve

eq_label <- paste0("y = ", format(curve_slope, digits = 4, scientific = TRUE), "x + ",
                    sprintf("%.4f", curve_intercept),
                    "\nR² = ", sprintf("%.4f", curve_r2))

standard_curve_plot <- ggplot(standards_clean, aes(x = std_conc_known, y = mean_595)) +
  geom_smooth(method = "lm", formula = y ~ x, se = FALSE, color = "steelblue", linewidth = 0.6) +
  geom_errorbar(aes(ymin = mean_595 - sd_595, ymax = mean_595 + sd_595),
                width = 30, color = "grey40") +
  geom_point(size = 2.6, color = "steelblue") +
  annotate("text", x = -Inf, y = Inf, label = eq_label,
           hjust = -0.08, vjust = 1.3, size = 3.6, fontface = "bold") +
  labs(x = "BSA standard concentration (µg/mL)",
       y = "Mean absorbance (595 nm)",
       title = "BSA standard curve",
       subtitle = "Points with technical-replicate CV(concentration) > 15% excluded (see below)") +
  theme_bw(base_size = 12)

print(standard_curve_plot)

ggsave(file.path(output_dir, "standard_curve_595.png"), standard_curve_plot,
       width = 7, height = 5.5, dpi = 300)

Standard points excluded from the curve:

excluded_standards <- group_stats %>% filter(type == "standard", high_cv)

cat("--- excluded_standards ---\n")
str(excluded_standards)
kable(excluded_standards %>%
        select(well_id, std_conc_known, mean_concentration, sd_concentration, cv_concentration),
      digits = 3)
--- excluded_standards ---
tibble [1 × 17] (S3: tbl_df/tbl/data.frame)
 $ well_id            : chr "STD7"
 $ name               : chr "BSA"
 $ type               : chr "standard"
 $ std_conc_known     : int 125
 $ base_sample_name   : chr NA
 $ dilution_code      : num NA
 $ dilution_multiplier: num NA
 $ n_wells            : int 3
 $ mean_595           : num 0.736
 $ sd_595             : num 0.031
 $ cv_595             : num 4.21
 $ mean_concentration : num 71.2
 $ sd_concentration   : num 59
 $ cv_concentration   : num 82.9
 $ n_wells_valid_conc : int 3
 $ is_zero_standard   : logi FALSE
 $ high_cv            : logi TRUE
well_id std_conc_known mean_concentration sd_concentration cv_concentration
STD7 125 71.197 58.999 82.867

9 APPLYING THE STANDARD CURVE TO SAMPLES

Samples that passed CV QC are back-calculated against the fitted curve to give a well-level concentration (calculated_concentration_well) – the concentration of whatever material was actually in the well, diluted or not. Multiplying by each sample’s dilution_multiplier then gives the final concentration (calculated_concentration) – the true concentration of the original, pre-dilution sample, which is the value reported in the results below. Samples whose mean 595 falls outside the retained standards’ absorbance range are flagged as extrapolated; this range check is done on the as-measured 595 value and is unaffected by dilution.

samples_clean <- clean_group_stats %>%
  filter(type == "sample") %>%
  mutate(
    se_595                         = sd_595 / sqrt(n_wells),
    calculated_concentration_well  = (mean_595 - curve_intercept) / curve_slope,
    calculated_concentration       = calculated_concentration_well * dilution_multiplier,
    in_curve_range                 = mean_595 >= min(standards_clean$mean_595) &
                                       mean_595 <= max(standards_clean$mean_595)
  ) %>%
  arrange(base_sample_name)

cat("--- samples_clean: CV-QC-passing samples, quantified against the fitted standard curve ---\n")
cat("    (calculated_concentration_well = as-measured; calculated_concentration = dilution-corrected)\n")
str(samples_clean)
--- samples_clean: CV-QC-passing samples, quantified against the fitted standard curve ---
    (calculated_concentration_well = as-measured; calculated_concentration = dilution-corrected)
tibble [9 × 21] (S3: tbl_df/tbl/data.frame)
 $ well_id                      : chr [1:9] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ name                         : chr [1:9] "F05_01_ambient-df.0" "F05_02_ambient-df.0" "F05_03_ambient-df.2" "F05_06_ambient-df.2" ...
 $ type                         : chr [1:9] "sample" "sample" "sample" "sample" ...
 $ std_conc_known               : int [1:9] NA NA NA NA NA NA NA NA NA
 $ base_sample_name             : chr [1:9] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ dilution_code                : num [1:9] 0 0 2 2 2 0 0 0 0
 $ dilution_multiplier          : num [1:9] 1 1 2 2 2 1 1 1 1
 $ n_wells                      : int [1:9] 3 3 3 3 3 3 3 3 3
 $ mean_595                     : num [1:9] 1.17 1.25 1.2 1.29 1.41 ...
 $ sd_595                       : num [1:9] 0.0267 0.03 0.0259 0.0375 0.0229 ...
 $ cv_595                       : num [1:9] 2.29 2.4 2.15 2.91 1.62 ...
 $ mean_concentration           : num [1:9] 884 1044 956 1120 1353 ...
 $ sd_concentration             : num [1:9] 50.1 57 48.4 71.7 43.2 ...
 $ cv_concentration             : num [1:9] 5.66 5.46 5.06 6.4 3.19 ...
 $ n_wells_valid_conc           : int [1:9] 3 3 3 3 3 3 3 3 3
 $ is_zero_standard             : logi [1:9] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ high_cv                      : logi [1:9] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ se_595                       : num [1:9] 0.0154 0.0173 0.0149 0.0217 0.0132 ...
 $ calculated_concentration_well: num [1:9] 876 1039 950 1116 1352 ...
 $ calculated_concentration     : num [1:9] 876 1039 1900 2232 2704 ...
 $ in_curve_range               : logi [1:9] TRUE TRUE TRUE TRUE TRUE TRUE ...
cat("Samples with absorbance outside the retained standard curve's range (extrapolated):\n")
kable(samples_clean %>% filter(!in_curve_range) %>%
        select(well_id, base_sample_name, mean_595, calculated_concentration),
      digits = 3)
Samples with absorbance outside the retained standard curve's range (extrapolated):

well_id base_sample_name mean_595 calculated_concentration ——— —————— ———- ————————–

9.1 Plot samples on standard curve

Every QC-passing sample is plotted at its own mean 595 and well-level back-calculated concentration (i.e. before dilution correction, since that is what is directly comparable to the curve’s own x/y relationship), with vertical error bars of ±1 standard error (of mean 595, across its 2-3 retained technical replicates), overlaid on the standard curve fit line and retained standard points.

samples_on_curve_plot <- ggplot() +
  geom_smooth(data = standards_clean, aes(x = std_conc_known, y = mean_595),
              method = "lm", formula = y ~ x, se = FALSE,
              color = "steelblue", linewidth = 0.6) +
  geom_point(data = standards_clean, aes(x = std_conc_known, y = mean_595),
             size = 2.6, color = "steelblue") +
  geom_errorbar(data = samples_clean,
                aes(x = calculated_concentration_well,
                    ymin = mean_595 - se_595, ymax = mean_595 + se_595),
                width = 30, color = "grey40") +
  geom_point(data = samples_clean,
             aes(x = calculated_concentration_well, y = mean_595, shape = in_curve_range),
             size = 2.4, color = "firebrick", alpha = 0.85) +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                      labels = c(`TRUE` = "in range", `FALSE` = "extrapolated"),
                      name = "Sample status") +
  labs(x = "Well-level concentration (µg/mL, pre-dilution-correction)",
       y = "Mean absorbance (595 nm)",
       title = "BSA standard curve with QC-passing samples overlaid",
       subtitle = "Samples plotted at their well-level back-calculated concentration; error bars = ±1 SE of mean 595") +
  theme_bw(base_size = 12)

print(samples_on_curve_plot)

ggsave(file.path(output_dir, "standard_curve_with_samples.png"), samples_on_curve_plot,
       width = 8, height = 5.5, dpi = 300)

10 RESULTS TABLE

results_table <- samples_clean %>%
  transmute(
    Sample                                   = base_sample_name,
    `Well ID`                                = well_id,
    `Dilution factor`                        = dilution_multiplier,
    `Mean 595`                               = round(mean_595, 3),
    `CV 595 (%)`                             = round(cv_595, 2),
    `CV concentration (%)`                   = round(cv_concentration, 2),
    `Well concentration (ug/mL)`             = round(calculated_concentration_well, 1),
    `Calculated concentration (ug/mL)`       = round(calculated_concentration, 1),
    `In standard curve range`                = ifelse(in_curve_range, "yes", "NO - extrapolated")
  ) %>%
  arrange(gsub("[0-9]+$", "", `Well ID`), as.numeric(gsub("^[A-Za-z]+", "", `Well ID`)))

kable(results_table)
Sample Well ID Dilution factor Mean 595 CV 595 (%) CV concentration (%) Well concentration (ug/mL) Calculated concentration (ug/mL) In standard curve range
F05_01_ambient SPL1 1 1.166 2.29 5.66 876.1 876.1 yes
F05_02_ambient SPL2 1 1.251 2.40 5.46 1038.9 1038.9 yes
F05_03_ambient SPL3 2 1.205 2.15 5.06 950.2 1900.4 yes
F05_06_ambient SPL4 2 1.291 2.91 6.40 1116.2 2232.4 yes
F05_08_ambient SPL5 2 1.414 1.62 3.19 1351.8 2703.7 yes
F07_05_36C SPL6 1 0.932 0.81 3.12 427.9 427.9 yes
F07_01_ambient SPL7 1 0.846 2.28 12.86 263.1 263.1 yes
F07_04_ambient SPL8 1 1.001 2.55 8.56 560.0 560.0 yes
F07_06_ambient SPL9 1 1.293 1.54 3.29 1119.4 1119.4 yes

Only samples passing all QC – CV ≤ 15% and within the retained standard curve’s range (i.e. not extrapolated) – are written to the CSV output.

results_table_export <- results_table %>% filter(`In standard curve range` == "yes")

cat("--- results_table_export: samples passing all QC (CV and in-range), written to CSV ---\n")
str(results_table_export)

write.csv(results_table_export, file.path(output_dir, "sample_protein_concentrations.csv"), row.names = FALSE)
--- results_table_export: samples passing all QC (CV and in-range), written to CSV ---
tibble [9 × 9] (S3: tbl_df/tbl/data.frame)
 $ Sample                          : chr [1:9] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ Well ID                         : chr [1:9] "SPL1" "SPL2" "SPL3" "SPL4" ...
 $ Dilution factor                 : num [1:9] 1 1 2 2 2 1 1 1 1
 $ Mean 595                        : num [1:9] 1.17 1.25 1.21 1.29 1.41 ...
 $ CV 595 (%)                      : num [1:9] 2.29 2.4 2.15 2.91 1.62 0.81 2.28 2.55 1.54
 $ CV concentration (%)            : num [1:9] 5.66 5.46 5.06 6.4 3.19 ...
 $ Well concentration (ug/mL)      : num [1:9] 876 1039 950 1116 1352 ...
 $ Calculated concentration (ug/mL): num [1:9] 876 1039 1900 2232 2704 ...
 $ In standard curve range         : chr [1:9] "yes" "yes" "yes" "yes" ...

11 SUMMARY

  • 17 sample/standard triplicate groups were imported (8 BSA standards, 9 samples).
  • 1 group(s) exceeded 15% technical-replicate CV(concentration), or had an undefined CV from insufficient valid replicates, and were excluded from all downstream analysis: STD7 (BSA).
  • The BSA standard curve was fit on the 7 remaining standard points: slope = 0.000522, intercept = 0.7086, R² = 0.9336.
  • 0 CV-QC-passing sample(s) fell outside the retained standard curve’s absorbance range and are flagged as extrapolated in the results table above.

11.1 Samples passing QC

Sample names are sorted naturally (ignoring stray punctuation) rather than by plain lexicographic string order. Concentrations shown are the final, dilution-corrected values.

natural_sort_key <- function(x) gsub("[^A-Za-z0-9]", "", x)
passing_samples_table <- samples_clean %>%
  transmute(sample_name = base_sample_name,
            `concentration(ug/mL)` = round(calculated_concentration, 1)) %>%
  arrange(natural_sort_key(sample_name))

cat("--- passing_samples_table: samples passing CV QC, name + calculated concentration ---\n")
str(passing_samples_table)

kable(passing_samples_table)
--- passing_samples_table: samples passing CV QC, name + calculated concentration ---
tibble [9 × 2] (S3: tbl_df/tbl/data.frame)
 $ sample_name         : chr [1:9] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_06_ambient" ...
 $ concentration(ug/mL): num [1:9] 876 1039 1900 2232 2704 ...
sample_name concentration(ug/mL)
F05_01_ambient 876.1
F05_02_ambient 1038.9
F05_03_ambient 1900.4
F05_06_ambient 2232.4
F05_08_ambient 2703.7
F07_01_ambient 263.1
F07_04_ambient 560.0
F07_05_36C 427.9
F07_06_ambient 1119.4
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase BSA https://robertslab.github.io/sams-notebook/posts/2026/2026-08-13-Protein-Quantification---June-2026-SORMI-M.gigas-Ctenidia-from-Famillies-5-and-7-for-Citrate-Synthase-Assay/ Thu, 13 Aug 2026 07:00:00 GMT Citrate Synthase Assay - SORMI June 2026 M.gigas Family 5 Ambient Ctenidia Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-11-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-5-Ambient-Ctenidia/

INTRO

Families 5 & 7 from the June 2026 SORMI experiment were selected for citrate synthase assays (selection by Steven). This is the first citrate synthase assay run on these samples: ctenidia from the eight Magallana gigas (Pacific oyster) Family 5 individuals held at ambient temperature, using the Citrate Synthase Assay Kit (ABCAM; cat: ab239712). Samples had been previously homogenized on 20260811 (Notebook entry), and protein concentrations for normalization were determined on 20260813 (Notebook entry).

Important

This experiment needs to be repeated. This notebook entry exists primarily for posterity.

The plate has QC failures (poor technical reps, primarily).

All data/code/analysis is available here:

MATERIALS & METHODS

Citrate Synthase Assay

The Citrate Synthase Assay Kit (ABCAM; cat: ab239712) was used according to the manufacturer’s instructions (ABCAM-Citrate-Synthase-Assay-v4a-ab239712.pdf (GitHub; PDF)).

Homogenates were thawed and 2 µL of undiluted homogenate was combined with 48 µL of Assay Buffer 7 per well, followed by 50 µL of Reaction Mix. Each sample was assayed in triplicate in a clear, flat-bottomed 96-well plate. Every sample also received a paired triplicate of background control wells, which receive Background Control Mix (no CS Substrate Mix) to correct for pre-existing free thiol in the homogenate. A six-point GSH standard curve (0, 8, 16, 24, 32, 40 nmol/well) and a CS positive control were included in triplicate on the same plate.

Master Mix Calcs

REACTION MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 43 50 2150
DTNB 5 50 250
CS Substrate 2 50 100

BACKGROUND MIX

REAGENT SINGLE_RXN_VOLUME (uL) REACTIONS TOTAL_VOLUME (uL)
Assay Buffer 7 45 30 1350
DTNB 5 30 150

Absorbance was read at 412 nm in kinetic mode at 25°C on a Synergy HTX plate reader (Agilent), at 2-minute intervals. This plate was read for 40 minutes (21 reads).

Normalization

Activity was normalized to total extracted protein rather than tissue weight. Tissue was homogenized in 350 µL of Assay Buffer 7, and protein concentration in each homogenate was quantified separately on 20260813 (Notebook entry) using the Bradford assay. Total protein mass per sample is that concentration multiplied by the 350 µL homogenization volume. Tissue weights are recorded in the plate layout labels for provenance only and play no role in the activity calculation.

Analysis

Data were analyzed in R. The Rmd file is here (GitHub):

Knitted markdown is below.

Data files

Plate reader files

Plate reader files (*.xpt) are available here (require Gen5 software to open):

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/plate_reader_files

Raw absorbance data

Raw absorbance data (*.csv), plate layouts, and Gen5 full reports are available here:

  • https://github.com/RobertsLab/sormi-assay-development/tree/main/Citrate_synthase/data/raw_absorbance

RESULTS

Four of eight samples are clean on every check — F05_01, F05_03, F05_07, and F05_08 — with no baseline-compromised replicates and technical CV < 8%. The instructive result is how little they differ: 7.18, 7.55, 8.62, and 6.77 mA412/min background-corrected, a spread of roughly 25% against a within-sample technical CV of 5-8%. Once F05_05 is recognized as a baseline artifact rather than a high-activity individual, no striking within-family variation remains at ambient temperature.

Because the standard curve failed its own QC, absolute activities carry uncertainty that the reported precision does not capture. Treat all values below as provisional.

Sample Reps used CV all reps (%) CV used reps (%) Corrected rate (mA412/min) Activity (mU/mg protein) Interpretation
F05_01_ambient 3/3 4.9 4.9 7.18 73.4 usable
F05_02_ambient 2/3 30.3 8.9 8.65 74.6 usable
F05_03_ambient 3/3 5.4 5.4 7.55 35.6 usable
F05_04_ambient 2/3 56.5 2.1 8.15 64.6 caution - some reps compromised
F05_05_ambient 2/3 68.8 15.1 19.10 240.4 DO NOT USE - no clean replicate
F05_06_ambient 2/3 36.8 24.7 6.95 27.9 caution - some reps compromised
F05_07_ambient 3/3 6.5 6.5 8.62 45.7 usable
F05_08_ambient 3/3 7.9 7.9 6.77 22.4 usable

Samples requiring re-assay: F05_05 (mandatory — no clean replicate) and F05_06 (24.7% CV on two genuinely disagreeing replicates). In practice the whole plate should be re-run, given the standard curve and background control problems.

AI Use Level 4: Substantial AI contribution

1 BACKGROUND

Citrate synthase (CS) activity in ctenidia of eight Magallana gigas (Pacific oyster) individuals from family F05 held at ambient temperature, assayed 2026-08-11 with the Abcam Citrate Synthase Assay Kit (ab239712), v4a.

The kit is a coupled kinetic assay. CS condenses acetyl-CoA and oxaloacetate, releasing free CoA-SH; the liberated thiol reduces DTNB to TNB2-, which absorbs at 412 nm. The rate of A412 increase is proportional to CS activity, and absolute nmol of thiol are assigned from a GSH (reduced glutathione) standard curve read on the same plate.

Absorbance was read at 412 nm in kinetic mode, 25 °C, every 2 min for 40 min (21 reads) on a Synergy HTX.

This document is fully self-contained: layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are all defined directly below (see ## Pipeline functions under SETUP) rather than sourced from a separate script, so the whole analysis can be reproduced from this one file.

1.1 Plate design

Plate rows Contents
A, B 8 samples (F05_01–F05_08), technical triplicates, + Reaction Mix
C, D Paired background controls for the same 8 samples, + Background Control Mix (no CS Substrate Mix)
E, F1–F6 GSH standards: 0, 8, 16, 24, 32, 40 nmol/well, triplicates
F7–F9 CS Positive Control, triplicate

Background control wells receive Background Control Mix, which omits CS Substrate Mix. They therefore contain no acetyl-CoA/oxaloacetate, cannot support the CS reaction, and should read flat and low for the whole kinetic run. Their purpose (Abcam §8.4 note, §10.2) is to correct for pre-existing free thiol / CoA in the homogenate, which produces a one-time DTNB reduction at t0 rather than an ongoing rate.

1.2 Sample naming

Following the convention in the sibling Glycogen assay (../../Glycogen/data/raw_luminescence/README.md), plate layout entries are:

  • <sample>-<assay_type>-<tissue_weight>-df.<dilution_factor>

where <sample> is itself composite:

  • <family>_<individual>_<temperature>, with _BG appended for background control wells.

E.g. F05_03_ambient-citrate_synthase-22.3-df.0 is family F05, individual 03, ambient exposure, citrate synthase assay, 22.3 mg of ctenidia tissue, undiluted homogenate. Standards follow STD-<assay_type>-<nmol_per_well> (e.g. STD-citrate_synthase-24 = 24 nmol GSH/well); the positive control is POS-citrate_synthase.

df.0 denotes no dilution of the homogenate, i.e. dilution factor D = 1 in the Abcam activity formula. It is not a literal multiplier of zero.

1.3 Important note(s)

  1. absorbance-*.csv is the sole data source. An earlier export of this file covered only plate columns 1–9 (54 of the 69 occupied wells); it has since been re-exported from the Gen5 software and now carries all 69 occupied wells (columns 1–12, rows A–F). The re-export was verified against full_report-*.txt — the two agree exactly on all 1,449 well × timepoint readings they share — and against layout-*.csv, which confirms the well set matches the plate layout with none missing and none extra. full_report-*.txt is retained only as a cross-check, not as a data source.
  2. Reaction volumes. 2 µL of homogenate + 48 µL Assay Buffer 7 per well, then 50 µL Reaction Mix. Sample volume V = 2 µL in the activity formula.
  3. Normalization is by total extracted protein, not tissue weight. Tissue was homogenized in 350 µL of Assay Buffer 7 (homogenate_volume_uL below), reported by the bench operator and applied to every sample. Protein concentration in each homogenate was measured separately (BCA/Bradford-style assay, 595 nm; see Protein concentration (normalization factor) below) and converted to a total protein mass by multiplying by this same 350 µL. Tissue input (10.1–30.6 mg, recorded in the well label for provenance only) is not used in any activity calculation on this plate — total extracted protein is the normalizer, because it reflects what the assay actually measures against (soluble protein carried into the 2 µL aliquot), not the starting tissue mass, which can vary in extraction efficiency from sample to sample.
  4. Rate windows are compared over a common early interval. Every trace on this plate decelerates — substrate is being consumed — so the measured rate depends on when it is measured, and a rate is only comparable between wells if the wells are measured over comparable intervals. Selecting each well’s steepest window independently satisfies this here because the steepest window is the initial one (0–8 min) for 22 of 24 sample wells; the two exceptions are wells whose traces are anomalous and are excluded anyway. The R2 floor for accepting a window is set to rate_min_r2 = 0.80 rather than a stricter value for the same reason: deceleration within the window depresses R2 even in clean wells (e.g. B10–B12 sit at 0.84–0.88 with 8% replicate CV), so a 0.95 floor would reject well-behaved replicates while retaining late, near-zero-rate windows from exhausted wells. Curvature is a property of the assay here, not a defect of the well.
  5. A412 at t = 0 is the master diagnostic on this plate. Clean sample wells start at A412 0.13–0.16. Every well that misbehaves — samples that decline, background wells that act like reactions — starts well above that. TNB present before the reaction begins means the well was contaminated or mis-dispensed, so its trace is not enzymology and a rate taken from it is not activity. sample_baseline_od = 0.35 separates the two populations cleanly (there are no sample wells between 0.23 and 0.29 except A6 at 0.29, itself anomalous). All three F05_05 replicates start elevated (0.65–0.93), so that sample has no clean replicate at all — see the results caveat below.
  6. Layout label typo, carried through from the plate layout file and handled explicitly in the parsing chunk: F05_05_ambient_BG wells (D1–D3) are labeled citrate_synthasse (three s’s).

2 SETUP

2.1 Libraries and shared pipeline functions

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Pipeline functions

All layout parsing, kinetic QC, standard-curve fitting, rate extraction, CV computation, protein normalization, and results assembly are defined directly in this document (rather than sourced from a separate script) so that it is fully self-contained and reproducible from this single file. The functions are grouped below by pipeline stage; each is called in the DATA / KINETIC TRACES / ANOMALY DETECTION / GSH STANDARD CURVE / RATE EXTRACTION / BACKGROUND CORRECTION / TECHNICAL REPLICATE PRECISION / CITRATE SYNTHASE ACTIVITY / QC SUMMARY sections that follow. Nothing plate-specific (typo fixes, disqualified samples, thresholds) is hardcoded in any function – it is always passed in as an argument, from the assay_params list defined further below or from a plate-specific fix list.

2.2.1 Data import and layout parsing

Reads the raw Gen5 exports (kinetic absorbance CSV, full-report text backup, plate-layout CSV), reconciles them against each other, and attaches well-type/sample-id/replicate metadata plus measured protein concentrations to every well.

## -----------------------------------------------------------------------
## elapsed_to_min()
## Purpose : Convert Gen5 "H:MM:SS" elapsed-time strings to decimal minutes.
## Inputs  : x - character vector of "H:MM:SS" strings
## Outputs : numeric vector of elapsed minutes, same length as x
## -----------------------------------------------------------------------
elapsed_to_min <- function(x) {
  vapply(strsplit(x, ":"), function(p) {
    p <- as.numeric(p)
    p[1] * 60 + p[2] + p[3] / 60
  }, numeric(1))
}

## -----------------------------------------------------------------------
## read_absorbance_csv()
## Purpose : Parse a Gen5 "absorbance-*.csv" kinetic export: wavelength
##           header, a "Time," row, a temperature row, then one row per well
##           (well ID in column 1, one OD reading per timepoint thereafter).
## Inputs  : path - path to the absorbance-*.csv file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_absorbance_csv <- function(path) {
  ln <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  time_i   <- grep("^Time,", ln)[1]
  time_str <- strsplit(ln[time_i], ",")[[1]]
  time_str <- time_str[!time_str %in% c("", "Time")]
  tmin     <- elapsed_to_min(time_str)

  well_ln <- ln[grepl("^[A-H][0-9]{1,2},", ln)]
  bind_rows(lapply(strsplit(well_ln, ","), function(p) {
    data.frame(well = p[1], time_min = tmin,
               od = as.numeric(p[seq_along(tmin) + 1]),
               stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## read_full_report()
## Purpose : Parse a Gen5 "full_report-*.txt" export: tab-delimited, wells
##           across columns, timepoints down rows. Used only as an
##           independent cross-check against the absorbance CSV, never as a
##           primary data source.
## Inputs  : path - path to the full_report-*.txt file
## Outputs : long-format data.frame, one row per well x timepoint:
##           well (chr), time_min (dbl), od (dbl)
## -----------------------------------------------------------------------
read_full_report <- function(path) {
  ln  <- iconv(readLines(path, warn = FALSE, encoding = "latin1"), "latin1", "UTF-8")
  hdr_i <- grep("^Time\t.*\tA1\t", ln)[1]
  hdr   <- trimws(strsplit(ln[hdr_i], "\t")[[1]])
  rows  <- ln[grepl("^[0-9]:[0-9]{2}:[0-9]{2}\t", ln)]
  m     <- do.call(rbind, strsplit(rows, "\t"))
  tmin  <- elapsed_to_min(m[, 1])

  well_cols <- which(grepl("^[A-H][0-9]{1,2}$", hdr))
  bind_rows(lapply(well_cols, function(j) {
    data.frame(well = hdr[j], time_min = tmin,
               od = as.numeric(m[, j]), stringsAsFactors = FALSE)
  }))
}

## -----------------------------------------------------------------------
## parse_plate_layout()
## Purpose : Convert a raw Gen5 plate-layout CSV export into one row per
##           occupied well with its descriptive label. Auto-detects two
##           export formats seen in practice:
##             "single"  - one row per plate row (row 1 = column numbers,
##                         column 1 = row letter, last column = "Name" tag);
##                         each cell already holds the full descriptive label.
##             "double"  - each plate row is exported as TWO rows: a
##                         "Well ID" row (SPL codes) immediately followed by
##                         a "Name" row (the descriptive label actually used
##                         downstream).
##           The format is detected from column 1: if a value repeats twice
##           in a row (once for the well-ID row, once for the label row)
##           immediately below a numeric header, it is treated as "double".
## Inputs  : path - path to the layout-*.csv file
## Outputs : data.frame, one row per occupied well:
##           well (chr, e.g. "A1"), plate_row (chr), plate_col (int),
##           label (chr, trimmed descriptive label)
## -----------------------------------------------------------------------
parse_plate_layout <- function(path) {
  plate_layout <- read.csv(path, header = FALSE, stringsAsFactors = FALSE)

  col1 <- trimws(plate_layout$V1)
  # "double" format: row letters appear twice in a row in column 1
  # (once tagging the Well-ID row, once tagging the Name row), e.g.
  # "A", "A", "B", "B", ... Detect by checking whether non-empty,
  # non-header values in col1 are each immediately followed by a repeat.
  data_rows   <- which(col1 != "" & !is.na(col1))[-1]  # drop header row (row 1)
  is_double <- length(data_rows) >= 2 &&
    all(col1[data_rows[c(TRUE, FALSE)]] == col1[data_rows[c(FALSE, TRUE)]][seq_len(floor(length(data_rows) / 2))])

  if (is_double) {
    # Pair each row-letter row with the following label row.
    row_letter_rows <- data_rows[c(TRUE, FALSE)]
    label_rows      <- data_rows[c(FALSE, TRUE)]
    stopifnot(length(row_letter_rows) == length(label_rows))

    out <- expand.grid(pair_idx = seq_along(row_letter_rows),
                        col_idx  = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = col1[row_letter_rows[pair_idx]],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j],
                                   label_rows[pair_idx], col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  } else {
    # "single" format: row 1 = column numbers, column 1 = row letters,
    # last column = "Name" tag; each cell already holds the full label.
    out <- expand.grid(row_idx = 2:nrow(plate_layout),
                        col_idx = 2:(ncol(plate_layout) - 1)) %>%
      mutate(
        plate_row = plate_layout$V1[row_idx],
        plate_col = as.integer(plate_layout[1, col_idx]),
        well      = paste0(plate_row, plate_col),
        label     = trimws(mapply(function(i, j) plate_layout[i, j], row_idx, col_idx))
      ) %>%
      filter(label != "") %>%
      select(well, plate_row, plate_col, label) %>%
      arrange(plate_row, plate_col)
  }

  attr(out, "layout_format") <- ifelse(is_double, "double", "single")
  out
}

## -----------------------------------------------------------------------
## reconcile_kinetic_sources()
## Purpose : Cross-check the absorbance CSV (primary data source) against
##           the full_report text export (independent instrument export)
##           and the plate layout (expected well set). Fails loudly
##           (stopifnot) if any layout well is missing from the CSV, or if
##           the two kinetic sources disagree on any shared reading.
## Inputs  : absorbance_csv - long-format data.frame from read_absorbance_csv()
##           full_report    - long-format data.frame from read_full_report()
##           layout_wells   - data.frame from parse_plate_layout()
##           tolerance      - max allowed |CSV - report| absolute difference
##                            (default 1e-9, i.e. exact agreement)
## Outputs : list(plate_readings, overlap_check, summary), where
##           plate_readings = absorbance_csv with a `source` column added,
##           summary        = named list of counts for reporting
## -----------------------------------------------------------------------
reconcile_kinetic_sources <- function(absorbance_csv, full_report, layout_wells,
                                       tolerance = 1e-9) {
  wells_csv    <- unique(absorbance_csv$well)
  wells_report <- unique(full_report$well)

  wells_report_not_csv <- setdiff(wells_report, wells_csv)
  wells_layout_not_csv <- setdiff(layout_wells$well, wells_csv)
  wells_csv_not_layout <- setdiff(wells_csv, layout_wells$well)

  overlap_check <- inner_join(absorbance_csv, full_report,
                              by = c("well", "time_min"), suffix = c("_csv", "_report")) %>%
    mutate(abs_diff = abs(od_csv - od_report))

  max_disagreement <- if (nrow(overlap_check) > 0) max(overlap_check$abs_diff) else NA_real_

  stopifnot(
    length(wells_layout_not_csv) == 0,
    is.na(max_disagreement) || max_disagreement < tolerance
  )

  plate_readings <- absorbance_csv %>% mutate(source = "absorbance_csv")

  list(
    plate_readings = plate_readings,
    overlap_check  = overlap_check,
    summary = list(
      n_wells_csv               = length(wells_csv),
      n_wells_report            = length(wells_report),
      n_wells_report_not_csv    = length(wells_report_not_csv),
      n_wells_layout_not_csv    = length(wells_layout_not_csv),
      n_wells_csv_not_layout    = length(wells_csv_not_layout),
      n_shared_readings         = nrow(overlap_check),
      max_disagreement          = max_disagreement
    )
  )
}

## -----------------------------------------------------------------------
## annotate_wells()
## Purpose : Join layout labels onto the kinetic readings and parse each
##           label into well_type, sample_id, family, individual,
##           temperature, tissue weight, and standard concentration.
##           Expected label grammar:
##             sample/background : <family>_<individual>_<temperature>[_BG]-<assay_type>-<weight>-df.<n>
##             standard           : STD-<assay_type>-<nmol_per_well>
##             positive control   : POS-<assay_type>
##           Label typos (e.g. a misspelled assay_type) are corrected before
##           parsing via `label_fixes`, a named character vector of
##           c(pattern = replacement) pairs applied in order with gsub().
## Inputs  : plate_readings - data.frame from reconcile_kinetic_sources()$plate_readings
##           layout_wells   - data.frame from parse_plate_layout()
##           label_fixes    - optional named character vector, e.g.
##                            c("citrate_synthasse" = "citrate_synthase")
## Outputs : plate_long - fully annotated long-format data.frame, one row
##           per well x timepoint, with well_type, sample_id, family,
##           individual, temperature, weight_mg, std_nmol columns added
## -----------------------------------------------------------------------
annotate_wells <- function(plate_readings, layout_wells, label_fixes = NULL) {
  plate_long <- plate_readings %>%
    left_join(layout_wells, by = "well") %>%
    mutate(
      well_type = case_when(
        grepl("^STD-", label) ~ "standard",
        grepl("^POS-", label) ~ "positive_control",
        grepl("_BG-",  label) ~ "background",
        TRUE                  ~ "sample"
      ),
      label_clean = label
    )

  if (!is.null(label_fixes)) {
    for (pat in names(label_fixes)) {
      plate_long$label_clean <- gsub(pat, label_fixes[[pat]], plate_long$label_clean)
    }
  }

  plate_long <- plate_long %>%
    mutate(
      sample_id   = ifelse(well_type %in% c("sample", "background"),
                           sub("(_BG)?-[a-zA-Z_]+-.*$", "", label_clean), NA_character_),
      family      = ifelse(!is.na(sample_id), sub("^(F[0-9]+)_.*$", "\\1", sample_id), NA_character_),
      individual  = ifelse(!is.na(sample_id), sub("^F[0-9]+_([0-9]+)_.*$", "\\1", sample_id), NA_character_),
      temperature = ifelse(!is.na(sample_id), sub("^F[0-9]+_[0-9]+_(.*)$", "\\1", sample_id), NA_character_),
      weight_mg   = ifelse(well_type %in% c("sample", "background"),
                           as.numeric(sub("^.*-[a-zA-Z_]+-([0-9.]+)-df\\..*$", "\\1", label_clean)),
                           NA_real_),
      std_nmol    = ifelse(well_type == "standard",
                           as.numeric(sub("^STD-[a-zA-Z_]+-", "", label_clean)), NA_real_)
    ) %>%
    arrange(well_type, sample_id, well, time_min)

  # Fail loudly rather than silently dropping malformed labels
  stopifnot(
    !any(is.na(plate_long$well_type)),
    !any(is.na(plate_long$weight_mg[plate_long$well_type %in% c("sample", "background")])),
    !any(is.na(plate_long$std_nmol[plate_long$well_type == "standard"]))
  )

  plate_long
}

## -----------------------------------------------------------------------
## load_protein_concentrations()
## Purpose : Load one or more BCA/Bradford-style protein-concentration CSV
##           exports, concatenate them, and match each plate sample to
##           exactly one protein-concentration record. Fails loudly if any
##           plate sample has zero or more than one match. Converts
##           concentration (ug/mL) to total protein mass (mg) in the full
##           homogenate using the measured homogenization volume.
## Inputs  : protein_files        - character vector of CSV paths, each with
##                                   columns `Sample` and
##                                   `Calculated concentration (ug/mL)`
##           plate_sample_ids     - character vector of this plate's sample IDs
##           homogenate_volume_uL - measured homogenization volume (uL),
##                                   applied identically to every sample
##           sample_id_fixes      - optional named character vector of
##                                   c(pattern = replacement) label-typo
##                                   corrections applied to the protein
##                                   file's Sample column before matching,
##                                   e.g. c("F-05_07_36C" = "F05_07_36C")
## Outputs : data.frame, one row per plate sample:
##           sample_id, source_file, conc_ug_mL, total_protein_mg
## -----------------------------------------------------------------------
load_protein_concentrations <- function(protein_files, plate_sample_ids,
                                         homogenate_volume_uL,
                                         sample_id_fixes = NULL) {
  protein_all <- protein_files %>%
    purrr::map_dfr(~ read.csv(.x, check.names = FALSE) %>%
                      mutate(source_file = basename(.x))) %>%
    rename(sample_id_protein = Sample, conc_ug_mL = `Calculated concentration (ug/mL)`) %>%
    select(sample_id_protein, source_file, conc_ug_mL)

  if (!is.null(sample_id_fixes)) {
    for (pat in names(sample_id_fixes)) {
      protein_all$sample_id_protein <- gsub(pat, sample_id_fixes[[pat]],
                                             protein_all$sample_id_protein)
    }
  }

  protein_matches <- protein_all %>% filter(sample_id_protein %in% plate_sample_ids)
  match_counts    <- protein_matches %>% count(sample_id_protein, name = "n_matches")

  stopifnot(
    all(plate_sample_ids %in% protein_matches$sample_id_protein),
    all(match_counts$n_matches == 1)
  )

  protein_matches %>%
    rename(sample_id = sample_id_protein) %>%
    mutate(total_protein_mg = conc_ug_mL * homogenate_volume_uL / 1e6) %>%
    arrange(sample_id)
}

2.2.2 Well-level kinetic QC

Flags read glitches, non-rising (‘decreasing’) traces, elevated t0 baselines, and other well-level anomalies before any rate is trusted.

## -----------------------------------------------------------------------
## local_step_excess()
## Purpose : For a kinetic trace, measure how much each read-to-read step
##           departs from the AVERAGE of its two immediate neighbouring
##           steps. A large local excess flags a read glitch; comparing to
##           local neighbours (rather than the trace's overall median step)
##           tolerates traces that smoothly decelerate throughout.
## Inputs  : od - numeric vector of OD readings in time order
## Outputs : numeric vector of length length(od) - 1 (one per step)
## -----------------------------------------------------------------------
local_step_excess <- function(od) {
  d <- diff(od)
  n <- length(d)
  local_expect <- vapply(seq_len(n), function(i) {
    mean(d[intersect(c(i - 1L, i + 1L), seq_len(n))])
  }, numeric(1))
  d - local_expect
}

## -----------------------------------------------------------------------
## compute_well_diagnostics()
## Purpose : Compute four independent per-well trace-shape diagnostics from
##           the raw kinetic trace, before any rate fitting: direction
##           (net change), monotonicity (fraction of rising steps),
##           over-range (any read above the photometric linearity ceiling),
##           and discontinuity (a read-to-read glitch via local_step_excess).
##           Also flags an elevated starting absorbance for sample wells,
##           and background wells that behave like active reactions.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with od_linear_max, glitch_excess_od,
##                          sample_baseline_od, bg_flat_od_max,
##                          bg_flat_drift_max
## Outputs : well_diagnostics - one row per well, with od_first, od_last,
##           od_max, net_change, frac_rising, max_step, typical_step,
##           step_excess, glitch_at_min, and flag_* / n_flags columns
## -----------------------------------------------------------------------
compute_well_diagnostics <- function(plate_long, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, plate_row, plate_col, well_type, sample_id, label, std_nmol, source) %>%
    summarise(
      od_first        = first(od),
      od_last         = last(od),
      od_max          = max(od),
      net_change      = last(od) - first(od),
      frac_rising     = mean(diff(od) > 0),
      max_step        = diff(od)[which.max(abs(diff(od)))],
      typical_step    = median(abs(diff(od))),
      step_excess     = local_step_excess(od)[which.max(abs(local_step_excess(od)))],
      glitch_at_min   = time_min[which.max(abs(local_step_excess(od))) + 1L],
      .groups = "drop"
    ) %>%
    mutate(
      step_ratio      = abs(max_step) / pmax(typical_step, 0.001),
      flag_decreasing = net_change <= 0,
      flag_over_range = od_max > assay_params$od_linear_max,
      flag_glitch     = abs(step_excess) > assay_params$glitch_excess_od,
      flag_high_baseline = well_type == "sample" & od_first > assay_params$sample_baseline_od,
      flag_bg_active  = well_type == "background" &
                        (od_first > assay_params$bg_flat_od_max |
                         abs(net_change) > assay_params$bg_flat_drift_max),
      n_flags         = flag_decreasing + flag_over_range + flag_glitch +
                        flag_bg_active + flag_high_baseline
    )
}

## -----------------------------------------------------------------------
## compute_baseline_diagnostics()
## Purpose : Classify sample/background wells by starting absorbance
##           (elevated vs. normal) and summarize, per sample, how many of
##           its replicates are baseline-compromised. A sample with all
##           replicates elevated has no clean replicate at all.
## Inputs  : well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with sample_baseline_od
## Outputs : list(baseline_check, baseline_per_sample)
## -----------------------------------------------------------------------
compute_baseline_diagnostics <- function(well_diagnostics, assay_params) {
  baseline_check <- well_diagnostics %>%
    filter(well_type %in% c("sample", "background")) %>%
    mutate(baseline = ifelse(od_first > assay_params$sample_baseline_od,
                             "elevated", "normal")) %>%
    arrange(desc(od_first))

  baseline_per_sample <- baseline_check %>%
    filter(well_type == "sample") %>%
    group_by(sample_id) %>%
    summarise(n_elevated = sum(baseline == "elevated"), n = n(),
              median_baseline = median(od_first), .groups = "drop") %>%
    arrange(desc(n_elevated))

  list(baseline_check = baseline_check, baseline_per_sample = baseline_per_sample)
}

2.2.3 Standard curve and rate extraction

Fits the GSH standard curve, extracts each well’s rate via a sliding-window linear regression gated on R-squared, and summarizes the positive control.

## -----------------------------------------------------------------------
## fit_standard_curve()
## Purpose : Extract t=0 GSH standard readings, quantify signal drift over
##           the run, compute per-concentration replicate statistics, flag
##           outlier wells by deviation from their triplicate median, and
##           fit three candidate calibration lines (all wells, per-
##           concentration means, outlier-excluded). The outlier-excluded
##           fit is the calibration used downstream.
## Inputs  : plate_long   - data.frame from annotate_wells()
##           assay_params - list with std_outlier_od
## Outputs : list with standards_t0, standard_drift, standard_summary,
##           standards_flagged, fit_all_wells, fit_means, fit_no_outliers,
##           fit_comparison, std_slope, std_intercept, std_r2, std_nmol_max
## -----------------------------------------------------------------------
fit_standard_curve <- function(plate_long, assay_params) {
  standards_all <- plate_long %>%
    filter(well_type == "standard") %>%
    select(well, std_nmol, time_min, od, source)

  standards_t0 <- standards_all %>%
    filter(time_min == 0) %>%
    select(well, std_nmol, od, source) %>%
    arrange(std_nmol, well)

  standard_drift <- standards_all %>%
    group_by(well, std_nmol) %>%
    summarise(od_t0 = od[time_min == 0], od_t40 = od[time_min == max(time_min)],
              drift = od_t40 - od_t0, .groups = "drop") %>%
    arrange(std_nmol, well)

  standard_summary <- standards_t0 %>%
    group_by(std_nmol) %>%
    summarise(n = n(), mean_od = mean(od), sd_od = sd(od),
              se_od = sd(od) / sqrt(n()), cv_pct = 100 * sd(od) / mean(od),
              median_od = median(od), .groups = "drop") %>%
    arrange(std_nmol) %>%
    mutate(net_od = mean_od - mean_od[std_nmol == 0],
           od_per_nmol = ifelse(std_nmol > 0, net_od / std_nmol, NA_real_))

  standards_flagged <- standards_t0 %>%
    group_by(std_nmol) %>%
    mutate(triplicate_median = median(od),
           deviation = od - triplicate_median,
           is_outlier = abs(deviation) > assay_params$std_outlier_od) %>%
    ungroup() %>%
    arrange(desc(abs(deviation)))

  fit_all_wells   <- lm(od ~ std_nmol, data = standards_flagged)
  fit_means       <- lm(mean_od ~ std_nmol, data = standard_summary)
  fit_no_outliers <- lm(od ~ std_nmol, data = standards_flagged %>% filter(!is_outlier))

  fit_comparison <- bind_rows(
    data.frame(fit = "all wells",          n = nobs(fit_all_wells),
               slope = coef(fit_all_wells)[2],   intercept = coef(fit_all_wells)[1],
               r_squared = summary(fit_all_wells)$r.squared),
    data.frame(fit = "concentration means", n = nobs(fit_means),
               slope = coef(fit_means)[2],       intercept = coef(fit_means)[1],
               r_squared = summary(fit_means)$r.squared),
    data.frame(fit = "outlier-excluded",   n = nobs(fit_no_outliers),
               slope = coef(fit_no_outliers)[2], intercept = coef(fit_no_outliers)[1],
               r_squared = summary(fit_no_outliers)$r.squared)
  )

  list(
    standards_all      = standards_all,
    standards_t0       = standards_t0,
    standard_drift     = standard_drift,
    standard_summary   = standard_summary,
    standards_flagged  = standards_flagged,
    fit_all_wells      = fit_all_wells,
    fit_means          = fit_means,
    fit_no_outliers    = fit_no_outliers,
    fit_comparison     = fit_comparison,
    std_slope          = unname(coef(fit_no_outliers)[2]),
    std_intercept      = unname(coef(fit_no_outliers)[1]),
    std_r2             = summary(fit_no_outliers)$r.squared,
    std_nmol_max       = max(standard_summary$std_nmol)
  )
}

## -----------------------------------------------------------------------
## window_slopes()
## Purpose : Fit every sliding window of `w` consecutive timepoints on one
##           kinetic trace and return the slope (mOD/min) and R^2 of each.
## Inputs  : t - numeric vector of timepoints (min)
##           y - numeric vector of OD readings, same length as t
##           w - window width in number of points
## Outputs : data.frame, one row per window: t_start, t_end,
##           slope_mOD_min, r2
## -----------------------------------------------------------------------
window_slopes <- function(t, y, w) {
  n <- length(y)
  bind_rows(lapply(seq_len(n - w + 1), function(s) {
    i  <- s:(s + w - 1)
    f  <- lm(y[i] ~ t[i])
    r2 <- summary(f)$r.squared
    data.frame(t_start = t[i[1]], t_end = t[i[w]],
               slope_mOD_min = unname(coef(f)[2]) * 1000,
               r2 = ifelse(is.na(r2), 0, r2))
  }))
}

## -----------------------------------------------------------------------
## compute_well_rates()
## Purpose : For every well, scan all sliding windows of assay_params$
##           rate_window_n consecutive reads and record the max-increasing
##           window (used for activity) and the max-absolute window
##           (diagnostic only -- matches Gen5's own `Max V` convention,
##           which can report large negative "rates" for degrading wells).
##           A rate is `rate_usable` only if its R^2 clears the floor, the
##           well's net change is positive, and no read glitch falls inside
##           the fitted window.
## Inputs  : plate_long       - data.frame from annotate_wells()
##           well_diagnostics - data.frame from compute_well_diagnostics()
##           assay_params     - list with rate_window_n, rate_min_r2
## Outputs : well_rates - one row per well: t_start, t_end, slope_mOD_min,
##           r2, max_abs_slope_mOD_min, abs_window_is_negative,
##           glitch_in_window, rate_usable (plus diagnostic columns joined in)
## -----------------------------------------------------------------------
compute_well_rates <- function(plate_long, well_diagnostics, assay_params) {
  plate_long %>%
    arrange(well, time_min) %>%
    group_by(well, well_type, sample_id, std_nmol) %>%
    group_modify(~ {
      w   <- window_slopes(.x$time_min, .x$od, assay_params$rate_window_n)
      inc <- w[which.max(w$slope_mOD_min), ]
      abs_ <- w[which.max(abs(w$slope_mOD_min)), ]
      data.frame(
        t_start = inc$t_start, t_end = inc$t_end,
        slope_mOD_min = inc$slope_mOD_min, r2 = inc$r2,
        max_abs_slope_mOD_min = abs_$slope_mOD_min,
        abs_window_is_negative = abs_$slope_mOD_min < 0
      )
    }) %>%
    ungroup() %>%
    left_join(well_diagnostics %>% select(well, net_change, frac_rising, od_max,
                                          flag_decreasing, flag_over_range,
                                          flag_glitch, glitch_at_min,
                                          flag_bg_active, n_flags),
              by = "well") %>%
    mutate(
      glitch_in_window = flag_glitch & glitch_at_min > t_start & glitch_at_min <= t_end,
      rate_usable      = r2 >= assay_params$rate_min_r2 & net_change > 0 & !glitch_in_window
    )
}

## -----------------------------------------------------------------------
## compute_positive_control()
## Purpose : Compute the CS Positive Control's rate and activity, and check
##           that its replicates rose linearly -- the plate's proof that the
##           reaction chemistry (DTNB, coupling, reader) worked, independent
##           of any sample-specific problem.
## Inputs  : well_rates   - data.frame from compute_well_rates()
##           std_slope    - calibration slope (A412/nmol), from fit_standard_curve()
##           assay_params - list with sample_volume_uL, dilution_factor
## Outputs : pos_control - one row per positive-control well: t_start, t_end,
##           slope_mOD_min, r2, net_change, activity_mU_uL
## -----------------------------------------------------------------------
compute_positive_control <- function(well_rates, std_slope, assay_params) {
  well_rates %>%
    filter(well_type == "positive_control") %>%
    mutate(activity_mU_uL = (slope_mOD_min / 1000) / std_slope /
                            assay_params$sample_volume_uL * assay_params$dilution_factor)
}

2.2.4 Background correction, replicate CV, and activity calculation

Estimates background rate from flat background wells, computes technical-replicate coefficients of variation, and converts corrected rates into protein-normalized CS activity.

## -----------------------------------------------------------------------
## compute_background_correction()
## Purpose : Estimate each sample's background rate from its flat
##           (well-behaved) background-control replicates only, since
##           anomalous background wells do not measure background.
## Inputs  : well_rates - data.frame from compute_well_rates()
## Outputs : background_per_sample - one row per sample: n_bg_total,
##           n_bg_flat, bg_rate_flat (NA if no flat replicate exists),
##           bg_rate_all
## -----------------------------------------------------------------------
compute_background_correction <- function(well_rates) {
  well_rates %>%
    filter(well_type == "background") %>%
    group_by(sample_id) %>%
    summarise(
      n_bg_total    = n(),
      n_bg_flat     = sum(!flag_bg_active),
      bg_rate_flat  = ifelse(sum(!flag_bg_active) > 0,
                             mean(slope_mOD_min[!flag_bg_active]), NA_real_),
      bg_rate_all   = mean(slope_mOD_min),
      .groups = "drop"
    )
}

## -----------------------------------------------------------------------
## compute_replicate_cv()
## Purpose : Compute technical-replicate coefficient of variation (CV%) on
##           the extracted rate, both across ALL three replicates and
##           across USABLE replicates only (rate_usable == TRUE). Flags any
##           sample whose CV exceeds assay_params$cv_threshold_pct.
## Inputs  : well_rates       - data.frame from compute_well_rates()
##           sample_id_order  - character vector giving the row order /
##                              complete sample-ID set to report (typically
##                              the plate's sample IDs, e.g. from
##                              protein_by_sample$sample_id)
##           assay_params     - list with cv_threshold_pct
## Outputs : cv_summary - one row per sample: n_all, mean_all, sd_all,
##           cv_all, n_usable, mean_usable, sd_usable, cv_usable,
##           excluded_wells, fails_cv_all, fails_cv_usable
## -----------------------------------------------------------------------
compute_replicate_cv <- function(well_rates, sample_id_order, assay_params) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  cv_all <- sample_rates %>%
    group_by(sample_id) %>%
    summarise(n_all = n(), mean_all = mean(slope_mOD_min), sd_all = sd(slope_mOD_min),
              cv_all = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  cv_usable <- sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_usable = n(), mean_usable = mean(slope_mOD_min),
              sd_usable = sd(slope_mOD_min),
              cv_usable = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop")

  data.frame(sample_id = sample_id_order, stringsAsFactors = FALSE) %>%
    left_join(cv_all, by = "sample_id") %>%
    left_join(cv_usable, by = "sample_id") %>%
    mutate(
      excluded_wells = vapply(sample_id, function(s) {
        w <- sample_rates$well[sample_rates$sample_id == s & !sample_rates$rate_usable]
        if (length(w) == 0) "-" else paste(w, collapse = ", ")
      }, character(1)),
      fails_cv_all    = cv_all > assay_params$cv_threshold_pct,
      fails_cv_usable = !is.na(cv_usable) & cv_usable > assay_params$cv_threshold_pct
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## calculate_cs_activity()
## Purpose : Compute per-sample CS activity following Abcam sec 10.3, using
##           only usable replicates, corrected for background rate, scaled
##           through the standard curve, and normalized to total extracted
##           protein (rather than tissue weight).
## Inputs  : well_rates            - data.frame from compute_well_rates()
##           protein_by_sample     - data.frame from load_protein_concentrations()
##           background_per_sample - data.frame from compute_background_correction()
##           std_slope             - calibration slope (A412/nmol), from fit_standard_curve()
##           std_nmol_max          - max calibrated GSH concentration, from fit_standard_curve()
##           assay_params          - list with sample_volume_uL, dilution_factor,
##                                    homogenate_volume_uL, read_duration_min
##           plate_long            - data.frame from annotate_wells(), used to
##                                    attach family/individual/temperature
## Outputs : cs_activity - one row per sample: n_reps_used, mean_rate_mOD_min,
##           sd_rate, cv_rate, family, individual, temperature, conc_ug_mL,
##           total_protein_mg, bg_rate_mOD_min, corrected_mOD_min,
##           activity_mU_per_uL, total_mU_in_homogenate,
##           activity_mU_per_mg_protein, within_std_range
## -----------------------------------------------------------------------
calculate_cs_activity <- function(well_rates, protein_by_sample, background_per_sample,
                                   std_slope, std_nmol_max, assay_params, plate_long) {
  sample_rates <- well_rates %>% filter(well_type == "sample") %>% arrange(sample_id, well)

  sample_rates %>%
    filter(rate_usable) %>%
    group_by(sample_id) %>%
    summarise(n_reps_used = n(), mean_rate_mOD_min = mean(slope_mOD_min),
              sd_rate = sd(slope_mOD_min),
              cv_rate = 100 * sd(slope_mOD_min) / mean(slope_mOD_min), .groups = "drop") %>%
    left_join(plate_long %>% filter(well_type == "sample") %>%
                distinct(sample_id, family, individual, temperature),
              by = "sample_id") %>%
    left_join(protein_by_sample %>% select(sample_id, conc_ug_mL, total_protein_mg),
              by = "sample_id") %>%
    left_join(background_per_sample %>% select(sample_id, bg_rate_flat, n_bg_flat),
              by = "sample_id") %>%
    mutate(
      bg_rate_mOD_min       = ifelse(is.na(bg_rate_flat), 0, bg_rate_flat),
      corrected_mOD_min     = mean_rate_mOD_min - bg_rate_mOD_min,
      rate_OD_min           = corrected_mOD_min / 1000,
      nmol_per_min          = rate_OD_min / std_slope,
      activity_mU_per_uL    = nmol_per_min / assay_params$sample_volume_uL *
                              assay_params$dilution_factor,
      total_mU_in_homogenate = activity_mU_per_uL * assay_params$homogenate_volume_uL,
      activity_mU_per_mg_protein = total_mU_in_homogenate / total_protein_mg,
      nmol_in_window        = nmol_per_min * (assay_params$read_duration_min),
      within_std_range      = nmol_in_window <= std_nmol_max
    ) %>%
    arrange(sample_id)
}

## -----------------------------------------------------------------------
## build_results_table()
## Purpose : Assemble the final, presentation-ready per-sample results
##           table from cs_activity, cv_summary and baseline diagnostics,
##           with a plain-language Interpretation column.
## Inputs  : cs_activity          - data.frame from calculate_cs_activity()
##           cv_summary           - data.frame from compute_replicate_cv()
##           baseline_per_sample  - data.frame from compute_baseline_diagnostics()
##           assay_params         - list with cv_threshold_pct, homogenate_volume_uL
## Outputs : results_table - formatted data.frame, ready for kable()/write.csv()
## -----------------------------------------------------------------------
build_results_table <- function(cs_activity, cv_summary, baseline_per_sample, assay_params) {
  cs_activity %>%
    left_join(cv_summary %>% select(sample_id, n_all, cv_all), by = "sample_id") %>%
    left_join(baseline_per_sample %>% select(sample_id, n_elevated), by = "sample_id") %>%
    transmute(
      Sample                       = sample_id,
      Family                       = family,
      Individual                   = individual,
      Temperature                  = temperature,
      `Protein conc (ug/mL)`       = round(conc_ug_mL, 1),
      `Total protein (mg)`         = round(total_protein_mg, 3),
      `Reps used`                  = paste0(n_reps_used, "/", n_all),
      `CV all reps (%)`            = round(cv_all, 1),
      `CV used reps (%)`           = round(cv_rate, 1),
      `Rate (mA412/min)`           = round(mean_rate_mOD_min, 2),
      `BG rate (mA412/min)`        = round(bg_rate_mOD_min, 3),
      `Corrected rate (mA412/min)` = round(corrected_mOD_min, 2),
      `Activity (mU/uL)`           = round(activity_mU_per_uL, 4),
      `Activity (mU/mg protein)`   = round(activity_mU_per_mg_protein, 3),
      `CV flag`                    = ifelse(cv_all > assay_params$cv_threshold_pct,
                                            paste0("FAIL >", assay_params$cv_threshold_pct, "%"), "pass"),
      `Elevated baseline reps`     = paste0(n_elevated, "/3"),
      Interpretation               = case_when(
        n_elevated == 3 ~ "DO NOT USE - no clean replicate",
        n_elevated > 0  ~ "caution - some reps compromised",
        cv_rate > assay_params$cv_threshold_pct | is.na(cv_rate) ~ "caution - imprecise",
        TRUE ~ "usable"
      )
    )
}

2.2.5 QC summary and narrative generation

Assembles the per-well QC summary table and auto-generates the plain-language QC findings bullets used in the QC SUMMARY section below.

## -----------------------------------------------------------------------
## build_qc_summary_table()
## Purpose : Consolidate every QC check performed across the pipeline into a
##           single two-column (check, value) table, suitable for kable() and
##           for writing to qc_summary.csv. Mirrors the individual find
##           functions but as compact, at-a-glance figures rather than prose.
## Inputs  : layout_wells, reconcile_summary, std_curve, well_diagnostics,
##           well_rates, cv_summary, baseline_per_sample, pos_control,
##           assay_params
## Outputs : qc_summary - data.frame with columns check, value (both chr)
## -----------------------------------------------------------------------
build_qc_summary_table <- function(layout_wells, reconcile_summary, std_curve,
                                    well_diagnostics, well_rates, cv_summary,
                                    baseline_per_sample, pos_control, assay_params) {
  standard_summary  <- std_curve$standard_summary
  standards_flagged <- std_curve$standards_flagged
  standard_drift    <- std_curve$standard_drift
  sample_rates      <- well_rates %>% filter(well_type == "sample")

  data.frame(
    check = c(
      "Occupied wells in layout",
      "Layout wells missing from absorbance CSV",
      "Max disagreement between absorbance CSV and full report",
      "Standard concentrations with replicate CV > threshold",
      "Standard wells flagged as outliers",
      "Standard curve R^2 (outlier-excluded, replicate level)",
      "Standard curve R^2 (all wells, replicate level)",
      "Standard wells that LOST signal over the run",
      "Standards above photometric linearity ceiling",
      "Background control wells behaving as active reactions",
      "Sample wells with a decreasing trace",
      "Sample wells with an elevated starting A412",
      "Samples with ALL THREE replicates baseline-compromised",
      "Sample wells with a read glitch",
      "Sample wells usable for rate extraction",
      "Samples with technical CV > threshold (all reps)",
      "Samples with technical CV > threshold (usable reps)",
      "Positive control replicates rising and linear"
    ),
    value = c(
      nrow(layout_wells),
      reconcile_summary$n_wells_layout_not_csv,
      sprintf("%.g", reconcile_summary$max_disagreement),
      paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
            collapse = ", "),
      paste0(sum(standards_flagged$is_outlier), "/", nrow(standards_flagged), " (",
             paste(standards_flagged$well[standards_flagged$is_outlier], collapse = ", "), ")"),
      sprintf("%.4f", std_curve$std_r2),
      sprintf("%.4f", summary(std_curve$fit_all_wells)$r.squared),
      paste0(sum(standard_drift$drift < 0), "/", nrow(standard_drift)),
      paste(standard_summary$std_nmol[standard_summary$mean_od > assay_params$od_linear_max],
            collapse = ", "),
      paste0(sum(well_diagnostics$flag_bg_active), "/",
             sum(well_diagnostics$well_type == "background")),
      paste0(sum(well_diagnostics$flag_decreasing & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(well_diagnostics$flag_high_baseline), "/",
             sum(well_diagnostics$well_type == "sample"), " (",
             paste(well_diagnostics$well[well_diagnostics$flag_high_baseline], collapse = ", "), ")"),
      paste0(sum(baseline_per_sample$n_elevated == baseline_per_sample$n), "/",
             nrow(baseline_per_sample), " (",
             paste(baseline_per_sample$sample_id[baseline_per_sample$n_elevated == baseline_per_sample$n],
                   collapse = ", "), ")"),
      paste0(sum(well_diagnostics$flag_glitch & well_diagnostics$well_type == "sample"), "/",
             sum(well_diagnostics$well_type == "sample")),
      paste0(sum(sample_rates$rate_usable), "/", nrow(sample_rates)),
      paste0(sum(cv_summary$fails_cv_all), "/", nrow(cv_summary)),
      paste0(sum(cv_summary$fails_cv_usable), "/", nrow(cv_summary)),
      paste0(sum(pos_control$net_change > 0 & pos_control$r2 > 0.99), "/", nrow(pos_control))
    ),
    stringsAsFactors = FALSE
  )
}

## -----------------------------------------------------------------------
## generate_qc_findings()
## Purpose : Walk the QC/CV/anomaly objects for a plate and emit a
##           consistent, generic-language markdown bullet list of findings
##           (decreasing/glitched/high-baseline wells, background-control
##           mix-ups, standard-curve QC failures, CV failures, disqualified
##           samples). Intended to seed the SUMMARY section of a plate's
##           .Rmd; a short hand-written subsection for plate-specific
##           interpretation should follow it.
## Inputs  : well_diagnostics    - data.frame from compute_well_diagnostics()
##           well_rates          - data.frame from compute_well_rates()
##           cv_summary          - data.frame from compute_replicate_cv()
##           baseline_per_sample - data.frame from compute_baseline_diagnostics()
##           std_curve           - list from fit_standard_curve()
##           reconcile_summary   - list from reconcile_kinetic_sources()$summary
##           assay_params        - list with cv_threshold_pct, od_linear_max
## Outputs : character vector, one markdown bullet per finding (empty
##           findings are omitted, so a clean plate returns a short list)
## -----------------------------------------------------------------------
generate_qc_findings <- function(well_diagnostics, well_rates, cv_summary,
                                  baseline_per_sample, std_curve,
                                  reconcile_summary, assay_params) {
  bullets <- character(0)

  # Source reconciliation
  if (reconcile_summary$n_wells_layout_not_csv > 0 ||
      (!is.na(reconcile_summary$max_disagreement) && reconcile_summary$max_disagreement >= 1e-9)) {
    bullets <- c(bullets, sprintf(
      "**Kinetic source disagreement.** %d layout well(s) missing from the absorbance CSV; max disagreement between absorbance CSV and full report was %.4g A412.",
      reconcile_summary$n_wells_layout_not_csv, reconcile_summary$max_disagreement))
  } else {
    bullets <- c(bullets, sprintf(
      "**Kinetic sources agree.** absorbance CSV and full report agree exactly on all %d shared well x timepoint readings; all layout wells are present.",
      reconcile_summary$n_shared_readings))
  }

  # Sample wells with decreasing kinetics
  dec_samples <- well_diagnostics %>% filter(well_type == "sample", flag_decreasing)
  if (nrow(dec_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Decreasing kinetics.** %d sample well(s) end lower than (or equal to) their starting A412 and are excluded from rate extraction: %s.",
      nrow(dec_samples), paste(dec_samples$well, collapse = ", ")))
  }

  # Read glitches
  glitch_samples <- well_diagnostics %>% filter(well_type == "sample", flag_glitch)
  if (nrow(glitch_samples) > 0) {
    bullets <- c(bullets, sprintf(
      "**Read glitches.** %d sample well(s) contain a single read-to-read step inconsistent with its neighbours: %s (only disqualifying if it falls inside the fitted rate window).",
      nrow(glitch_samples), paste(glitch_samples$well, collapse = ", ")))
  }

  # Elevated baseline
  high_baseline <- well_diagnostics %>% filter(well_type == "sample", flag_high_baseline)
  if (nrow(high_baseline) > 0) {
    bullets <- c(bullets, sprintf(
      "**Elevated starting absorbance.** %d sample well(s) start above the baseline threshold (A412 > %s), indicating pre-existing thiol/contamination before the reaction began: %s.",
      nrow(high_baseline), assay_params$sample_baseline_od, paste(high_baseline$well, collapse = ", ")))
  }

  # Samples fully disqualified by baseline
  disqualified <- baseline_per_sample %>% filter(n_elevated == n)
  if (nrow(disqualified) > 0) {
    bullets <- c(bullets, sprintf(
      "**Sample(s) with no clean replicate.** Every replicate of %s has an elevated starting A412; these samples have no usable measurement on this plate regardless of CV.",
      paste(disqualified$sample_id, collapse = ", ")))
  }

  # Over-range wells
  over_range <- well_diagnostics %>% filter(flag_over_range)
  if (nrow(over_range) > 0) {
    bullets <- c(bullets, sprintf(
      "**Over-range wells.** %d well(s) exceed the photometric linearity ceiling (A412 > %s): %s.",
      nrow(over_range), assay_params$od_linear_max, paste(over_range$well, collapse = ", ")))
  }

  # Background wells behaving as active reactions
  bg_active <- well_diagnostics %>% filter(well_type == "background", flag_bg_active)
  if (nrow(bg_active) > 0) {
    bullets <- c(bullets, sprintf(
      "**Background control wells behaving as active reactions.** %d of %d background wells show a starting A412 or drift inconsistent with a flat, inactive well: %s.",
      nrow(bg_active), sum(well_diagnostics$well_type == "background"),
      paste(bg_active$well, collapse = ", ")))
  }

  # Standard curve QC
  std_summary <- std_curve$standard_summary
  cv_fail_std <- std_summary %>% filter(cv_pct > assay_params$cv_threshold_pct)
  if (nrow(cv_fail_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**GSH standard curve replicate imprecision.** %d of %d concentrations exceed %s%% CV: %s nmol/well.",
      nrow(cv_fail_std), nrow(std_summary), assay_params$cv_threshold_pct,
      paste(cv_fail_std$std_nmol, collapse = ", ")))
  }
  n_std_outliers <- sum(std_curve$standards_flagged$is_outlier)
  if (n_std_outliers > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard outlier wells.** %d of %d standard wells deviate from their triplicate median by more than %s A412: %s.",
      n_std_outliers, nrow(std_curve$standards_flagged), assay_params$std_outlier_od,
      paste(std_curve$standards_flagged$well[std_curve$standards_flagged$is_outlier], collapse = ", ")))
  }
  over_range_std <- std_summary %>% filter(mean_od > assay_params$od_linear_max)
  if (nrow(over_range_std) > 0) {
    bullets <- c(bullets, sprintf(
      "**Standards above the photometric linear range.** %s nmol/well read above A412 %s; the calibration is trustworthy only below these concentrations.",
      paste(over_range_std$std_nmol, collapse = ", "), assay_params$od_linear_max))
  }
  n_drift_neg <- sum(std_curve$standard_drift$drift < 0)
  if (n_drift_neg > 0) {
    bullets <- c(bullets, sprintf(
      "**Standard signal decay.** %d of %d standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.",
      n_drift_neg, nrow(std_curve$standard_drift)))
  }
  bullets <- c(bullets, sprintf(
    "**Standard curve fit (outlier-excluded):** slope = %.5f A412/nmol, R^2 = %.4f (all-wells R^2 = %.4f).",
    std_curve$std_slope, std_curve$std_r2, summary(std_curve$fit_all_wells)$r.squared))

  # Technical replicate CV failures
  fail_all <- cv_summary %>% filter(fails_cv_all)
  if (nrow(fail_all) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% (all replicates).** %d of %d samples: %s.",
      assay_params$cv_threshold_pct, nrow(fail_all), nrow(cv_summary),
      paste(sprintf("%s (%.0f%%)", fail_all$sample_id, fail_all$cv_all), collapse = ", ")))
  } else {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV.** All %d samples are within %s%% CV across all three replicates.",
      nrow(cv_summary), assay_params$cv_threshold_pct))
  }
  fail_usable <- cv_summary %>% filter(fails_cv_usable)
  if (nrow(fail_usable) > 0) {
    bullets <- c(bullets, sprintf(
      "**Technical replicate CV > %s%% persists after excluding flagged replicates.** %s.",
      assay_params$cv_threshold_pct,
      paste(sprintf("%s (%.0f%%, n=%d)", fail_usable$sample_id, fail_usable$cv_usable,
                    fail_usable$n_usable), collapse = ", ")))
  }

  bullets
}

2.3 Assay parameters

All assay constants and QC thresholds are collected here so nothing downstream hard-codes a number.

assay_params <- list(
  sample_volume_uL      = 2,     # homogenate volume per reaction well (V in Abcam formula)
  buffer_volume_uL      = 48,    # Assay Buffer 7 added to reach 50 uL pre-Reaction Mix
  dilution_factor       = 1,     # D; layout `df.0` = undiluted homogenate
  homogenate_volume_uL  = 350,   # MEASURED buffer volume tissue was homogenized in (see note 3)
  read_wavelength_nm    = 412,
  read_interval_min     = 2,
  read_duration_min     = 40,
  rate_window_n         = 5,     # points per sliding regression window (5 pts = 8 min)
  rate_min_r2           = 0.80,  # min R^2 for a rate window to be trusted (see note 4)
  glitch_excess_od      = 0.02,  # |step - mean of neighbouring steps| flagging a read glitch
  cv_threshold_pct      = 15,    # technical-replicate CV QC threshold
  od_linear_max         = 1.5,   # upper A412 bound for reliable photometry
  std_outlier_od        = 0.15,  # |deviation from triplicate median| flagging a standard well
  sample_baseline_od    = 0.35,  # max acceptable t0 A412 for a sample well (see note 5)
  bg_flat_od_max        = 0.15,  # a well-behaved background well starts below this A412
  bg_flat_drift_max     = 0.05   # ...and drifts less than this over the whole run
)

cat("--- assay_params: assay constants and QC thresholds ---\n\n")
str(assay_params)
--- assay_params: assay constants and QC thresholds ---

List of 16
 $ sample_volume_uL    : num 2
 $ buffer_volume_uL    : num 48
 $ dilution_factor     : num 1
 $ homogenate_volume_uL: num 350
 $ read_wavelength_nm  : num 412
 $ read_interval_min   : num 2
 $ read_duration_min   : num 40
 $ rate_window_n       : num 5
 $ rate_min_r2         : num 0.8
 $ glitch_excess_od    : num 0.02
 $ cv_threshold_pct    : num 15
 $ od_linear_max       : num 1.5
 $ std_outlier_od      : num 0.15
 $ sample_baseline_od  : num 0.35
 $ bg_flat_od_max      : num 0.15
 $ bg_flat_drift_max   : num 0.05

2.4 Output directory

# Output directory for this analysis (matches this file's name, per ../code/README.md)
output_dir <- "../outputs/Gen5-20260811-mgig-citrate_synthase-F05-ambient"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- output_dir: destination for all figures and tables ---\n")
str(output_dir)
--- output_dir: destination for all figures and tables ---
 chr "../outputs/Gen5-20260811-mgig-citrate_synthase-F05-ambient"

3 DATA

Data are read from the local repo (../data/raw_absorbance/) so this document renders before/after the files are pushed to GitHub.

3.1 Plate layout

data_dir <- "../data/raw_absorbance"
run_stem <- "Gen5-20260811-mgig-citrate_synthase-F05-ambient"

layout_wells <- parse_plate_layout(file.path(data_dir, paste0("layout-", run_stem, ".csv")))

cat("Layout format detected:", attr(layout_wells, "layout_format"), "\n")
cat("Occupied wells in layout:", nrow(layout_wells), "\n")

cat("\n--- layout_wells: one row per occupied well, with its descriptive label ---\n\n")
str(layout_wells)
Layout format detected: single 
Occupied wells in layout: 69 

--- layout_wells: one row per occupied well, with its descriptive label ---

'data.frame':   69 obs. of  4 variables:
 $ well     : chr  "A1" "A2" "A3" "A4" ...
 $ plate_row: chr  "A" "A" "A" "A" ...
 $ plate_col: int  1 2 3 4 5 6 7 8 9 10 ...
 $ label    : chr  "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_02_ambient-citrate_synthase-17.5-df.0" ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "row_idx" "col_idx"
  ..$ dimnames:List of 2
  .. ..$ row_idx: chr [1:8] "row_idx=2" "row_idx=3" "row_idx=4" "row_idx=5" ...
  .. ..$ col_idx: chr [1:12] "col_idx= 2" "col_idx= 3" "col_idx= 4" "col_idx= 5" ...
 - attr(*, "layout_format")= chr "single"

3.2 Kinetic readers

absorbance_csv <- read_absorbance_csv(file.path(data_dir, paste0("absorbance-", run_stem, ".csv")))
full_report    <- read_full_report(file.path(data_dir, paste0("full_report-", run_stem, ".txt")))

cat("absorbance CSV : ", dplyr::n_distinct(absorbance_csv$well), " wells x ",
    dplyr::n_distinct(absorbance_csv$time_min), " timepoints\n", sep = "")
cat("full report    : ", dplyr::n_distinct(full_report$well), " wells x ",
    dplyr::n_distinct(full_report$time_min), " timepoints\n", sep = "")

cat("\n--- absorbance_csv: long-format readings from absorbance-*.csv ---\n\n")
str(absorbance_csv)
cat("\n--- full_report: long-format readings from full_report-*.txt ---\n\n")
str(full_report)
absorbance CSV : 69 wells x 21 timepoints
full report    : 69 wells x 21 timepoints

--- absorbance_csv: long-format readings from absorbance-*.csv ---

'data.frame':   1449 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.126 0.143 0.16 0.176 0.191 0.203 0.213 0.222 0.229 0.236 ...

--- full_report: long-format readings from full_report-*.txt ---

'data.frame':   1449 obs. of  3 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.126 0.143 0.16 0.176 0.191 0.203 0.213 0.222 0.229 0.236 ...

3.3 Cross-check against the full report

absorbance-*.csv is used for every well; no gap-fill is needed now that the re-exported file covers the full plate. full_report-*.txt is used only to confirm the re-export is trustworthy: every well it reports should also be in the CSV, every well the layout expects should be in the CSV, and every reading the two files share should agree exactly.

recon <- reconcile_kinetic_sources(absorbance_csv, full_report, layout_wells)
plate_readings <- recon$plate_readings

cat("Wells in absorbance CSV          :", recon$summary$n_wells_csv, "\n")
cat("Wells in full report             :", recon$summary$n_wells_report, "\n")
cat("Wells in full report not in CSV  :", recon$summary$n_wells_report_not_csv, "\n")
cat("Layout wells missing from CSV    :", recon$summary$n_wells_layout_not_csv, "\n")
cat("CSV wells not in layout          :", recon$summary$n_wells_csv_not_layout, "\n")
cat("Shared well x timepoints         :", recon$summary$n_shared_readings, "\n")
cat("Max |CSV - report| disagreement  :", recon$summary$max_disagreement, "\n")

cat("\n--- recon$overlap_check: per-reading comparison of the two raw files ---\n\n")
str(recon$overlap_check)
cat("\n--- plate_readings: readings for all occupied wells, from absorbance_csv ---\n\n")
str(plate_readings)
Wells in absorbance CSV          : 69 
Wells in full report             : 69 
Wells in full report not in CSV  : 0 
Layout wells missing from CSV    : 0 
CSV wells not in layout          : 0 
Shared well x timepoints         : 1449 
Max |CSV - report| disagreement  : 0 

--- recon$overlap_check: per-reading comparison of the two raw files ---

'data.frame':   1449 obs. of  5 variables:
 $ well     : chr  "A1" "A1" "A1" "A1" ...
 $ time_min : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od_csv   : num  0.126 0.143 0.16 0.176 0.191 0.203 0.213 0.222 0.229 0.236 ...
 $ od_report: num  0.126 0.143 0.16 0.176 0.191 0.203 0.213 0.222 0.229 0.236 ...
 $ abs_diff : num  0 0 0 0 0 0 0 0 0 0 ...

--- plate_readings: readings for all occupied wells, from absorbance_csv ---

'data.frame':   1449 obs. of  4 variables:
 $ well    : chr  "A1" "A1" "A1" "A1" ...
 $ time_min: num  0 2 4 6 8 10 12 14 16 18 ...
 $ od      : num  0.126 0.143 0.16 0.176 0.191 0.203 0.213 0.222 0.229 0.236 ...
 $ source  : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...

3.4 Annotate wells and parse metadata

# Tolerate the `citrate_synthasse` typo in the D1-D3 labels (see note 6)
plate_long <- annotate_wells(plate_readings, layout_wells,
                              label_fixes = c("citrate_synthasse" = "citrate_synthase"))

cat("Wells per type:\n")
print(plate_long %>% distinct(well, well_type) %>% count(well_type, name = "n_wells") %>%
        as.data.frame(), row.names = FALSE)
cat("\nSamples:", paste(sort(unique(na.omit(plate_long$sample_id))), collapse = ", "), "\n")
cat("Standards (nmol/well):", paste(sort(unique(na.omit(plate_long$std_nmol))), collapse = ", "), "\n")

cat("\n--- plate_long: fully annotated long-format plate, one row per well x timepoint ---\n\n")
str(plate_long)
Wells per type:
        well_type n_wells
       background      24
 positive_control       3
           sample      24
         standard      18

Samples: F05_01_ambient, F05_02_ambient, F05_03_ambient, F05_04_ambient, F05_05_ambient, F05_06_ambient, F05_07_ambient, F05_08_ambient 
Standards (nmol/well): 0, 8, 16, 24, 32, 40 

--- plate_long: fully annotated long-format plate, one row per well x timepoint ---

'data.frame':   1449 obs. of  15 variables:
 $ well       : chr  "C1" "C1" "C1" "C1" ...
 $ time_min   : num  0 2 4 6 8 10 12 14 16 18 ...
 $ od         : num  0.057 0.058 0.058 0.058 0.058 0.059 0.059 0.059 0.06 0.06 ...
 $ source     : chr  "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ plate_row  : chr  "C" "C" "C" "C" ...
 $ plate_col  : int  1 1 1 1 1 1 1 1 1 1 ...
 $ label      : chr  "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" ...
 $ well_type  : chr  "background" "background" "background" "background" ...
 $ label_clean: chr  "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" ...
 $ sample_id  : chr  "F05_01_ambient" "F05_01_ambient" "F05_01_ambient" "F05_01_ambient" ...
 $ family     : chr  "F05" "F05" "F05" "F05" ...
 $ individual : chr  "01" "01" "01" "01" ...
 $ temperature: chr  "ambient" "ambient" "ambient" "ambient" ...
 $ weight_mg  : num  10.1 10.1 10.1 10.1 10.1 10.1 10.1 10.1 10.1 10.1 ...
 $ std_nmol   : num  NA NA NA NA NA NA NA NA NA NA ...

3.5 Protein concentration (normalization factor)

Activity is normalized to total extracted protein rather than tissue weight (note 3). Protein concentration for each sample was measured on a separate BCA/Bradford-style plate (595 nm) and supplied as two CSV exports covering different subsets of samples across families F05 and F07; only the F05 ambient samples analyzed on this plate are relevant here, so the two files are concatenated and then filtered down to this plate’s 8 sample IDs.

protein_files <- c(
  "../data/BSA/raw_absorbance/sample_protein_concentrations_1.csv",
  "../data/BSA/raw_absorbance/sample_protein_concentrations_2.csv"
)

plate_sample_ids <- sort(unique(na.omit(plate_long$sample_id)))

protein_by_sample <- load_protein_concentrations(protein_files, plate_sample_ids,
                                                  assay_params$homogenate_volume_uL) %>%
  left_join(plate_long %>% filter(well_type == "sample") %>%
              distinct(sample_id, family, individual, temperature),
            by = "sample_id")

cat("Samples matched to a protein concentration record:", nrow(protein_by_sample),
    "/", length(plate_sample_ids), "\n")
cat("Protein concentration range (ug/mL):",
    paste(range(protein_by_sample$conc_ug_mL), collapse = " - "), "\n")
cat("Total protein per homogenate range (mg):",
    paste(round(range(protein_by_sample$total_protein_mg), 3), collapse = " - "), "\n\n")
print(protein_by_sample %>% select(sample_id, source_file, conc_ug_mL, total_protein_mg) %>%
        as.data.frame(), row.names = FALSE)

cat("\n--- protein_by_sample: matched protein concentration and total protein per homogenate ---\n\n")
str(protein_by_sample)
Samples matched to a protein concentration record: 8 / 8 
Protein concentration range (ug/mL): 711.7 - 2703.7 
Total protein per homogenate range (mg): 0.249 - 0.946 

      sample_id                         source_file conc_ug_mL total_protein_mg
 F05_01_ambient sample_protein_concentrations_2.csv      876.1         0.306635
 F05_02_ambient sample_protein_concentrations_2.csv     1038.9         0.363615
 F05_03_ambient sample_protein_concentrations_2.csv     1900.4         0.665140
 F05_04_ambient sample_protein_concentrations_1.csv     1129.4         0.395290
 F05_05_ambient sample_protein_concentrations_1.csv      711.7         0.249095
 F05_06_ambient sample_protein_concentrations_2.csv     2232.4         0.781340
 F05_07_ambient sample_protein_concentrations_1.csv     1687.8         0.590730
 F05_08_ambient sample_protein_concentrations_2.csv     2703.7         0.946295

--- protein_by_sample: matched protein concentration and total protein per homogenate ---

'data.frame':   8 obs. of  7 variables:
 $ sample_id       : chr  "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ source_file     : chr  "sample_protein_concentrations_2.csv" "sample_protein_concentrations_2.csv" "sample_protein_concentrations_2.csv" "sample_protein_concentrations_1.csv" ...
 $ conc_ug_mL      : num  876 1039 1900 1129 712 ...
 $ total_protein_mg: num  0.307 0.364 0.665 0.395 0.249 ...
 $ family          : chr  "F05" "F05" "F05" "F05" ...
 $ individual      : chr  "01" "02" "03" "04" ...
 $ temperature     : chr  "ambient" "ambient" "ambient" "ambient" ...

4 KINETIC TRACES

Every trace is plotted before any rate is extracted. In this assay the shape of the trace is the primary diagnostic: CS activity must produce a monotonically rising A412, and only a genuinely linear stretch may be used for a rate.

trace_plot <- ggplot(plate_long, aes(x = time_min, y = od, group = well, colour = well_type)) +
  geom_line(linewidth = 0.5) +
  geom_hline(yintercept = assay_params$od_linear_max,
             linetype = "dotted", colour = "grey40") +
  facet_wrap(~ plate_row + plate_col, ncol = 12,
             labeller = function(d) list(paste0(d$plate_row, d$plate_col))) +
  scale_colour_manual(values = c(sample = "#08519c", background = "#a63603",
                                 standard = "#238b45", positive_control = "#6a51a3")) +
  labs(title = "A412 kinetic traces, all occupied wells",
       subtitle = paste0("Dotted line = assumed photometric linearity ceiling (A412 = ",
                         assay_params$od_linear_max,
                         "). Rising traces only are usable for rate extraction."),
       x = "Time (min)", y = "A412", colour = "Well type") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        strip.text = element_text(size = 6),
        axis.text  = element_text(size = 5),
        legend.position = "bottom")

ggsave(file.path(output_dir, "kinetic_traces_all_wells.png"), trace_plot,
       width = 11, height = 8, dpi = 300)

cat("--- trace_plot: ggplot object structure ---\n\n")
summary(trace_plot)

trace_plot

--- trace_plot: ggplot object structure ---

data: well, time_min, od, source, plate_row, plate_col, label,
  well_type, label_clean, sample_id, family, individual, temperature,
  weight_mg, std_nmol [1449x15]
mapping:  x = ~time_min, y = ~od, group = ~well, colour = ~well_type
scales:   colour 
faceting:  ~plate_row, ~plate_col 
-----------------------------------
geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

5 ANOMALY DETECTION

5.1 Per-well trace diagnostics

Four independent checks per well, all computed from the raw trace before any rate fitting:

  1. Direction — net A412 change from first to last read. A CS-containing well must be positive.
  2. Monotonicity — fraction of the 20 read-to-read intervals that rise.
  3. Over-range — any read above the photometric linearity ceiling.
  4. Discontinuity — a single read-to-read step that departs from its two neighbouring steps, catching read glitches rather than biology. Comparing each step against its immediate neighbours (rather than against the well’s overall median step) matters here because every trace decelerates: the first interval of a decelerating trace is legitimately the largest of the run and a median-based rule flags it spuriously.
well_diagnostics <- compute_well_diagnostics(plate_long, assay_params)

cat("Flag counts across all", nrow(well_diagnostics), "wells:\n")
print(well_diagnostics %>%
        summarise(decreasing = sum(flag_decreasing), over_range = sum(flag_over_range),
                  glitch = sum(flag_glitch), bg_active = sum(flag_bg_active),
                  high_baseline = sum(flag_high_baseline),
                  clean = sum(n_flags == 0)) %>% as.data.frame(), row.names = FALSE)

# The starting A412 separates clean wells from compromised ones across every
# well type, so it is tabulated explicitly rather than only used as a flag
cat("\nStarting A412 (t = 0) by well type:\n")
print(well_diagnostics %>% group_by(well_type) %>%
        summarise(n = n(), min = min(od_first), median = median(od_first),
                  max = max(od_first), .groups = "drop") %>%
        as.data.frame(), row.names = FALSE, digits = 3)

cat("\n--- well_diagnostics: per-well trace shape diagnostics and anomaly flags ---\n\n")
str(well_diagnostics)
Flag counts across all 69 wells:
 decreasing over_range glitch bg_active high_baseline clean
         28          7     11         7             5    37

Starting A412 (t = 0) by well type:
        well_type  n   min median   max
       background 24 0.050  0.061 0.817
 positive_control  3 0.216  0.229 0.236
           sample 24 0.125  0.158 0.929
         standard 18 0.095  1.285 2.189

--- well_diagnostics: per-well trace shape diagnostics and anomaly flags ---

tibble [69 × 24] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:69] "A1" "A10" "A11" "A12" ...
 $ plate_row         : chr [1:69] "A" "A" "A" "A" ...
 $ plate_col         : int [1:69] 1 10 11 12 2 3 4 5 6 7 ...
 $ well_type         : chr [1:69] "sample" "sample" "sample" "sample" ...
 $ sample_id         : chr [1:69] "F05_01_ambient" "F05_04_ambient" "F05_04_ambient" "F05_04_ambient" ...
 $ label             : chr [1:69] "F05_01_ambient-citrate_synthase-10.1-df.0" "F05_04_ambient-citrate_synthase-14.0-df.0" "F05_04_ambient-citrate_synthase-14.0-df.0" "F05_04_ambient-citrate_synthase-14.0-df.0" ...
 $ std_nmol          : num [1:69] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:69] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:69] 0.126 0.129 0.136 0.671 0.129 0.125 0.135 0.154 0.287 0.154 ...
 $ od_last           : num [1:69] 0.29 0.286 0.292 0.331 0.291 0.294 0.297 0.317 0.328 0.292 ...
 $ od_max            : num [1:69] 0.29 0.286 0.292 0.671 0.291 0.294 0.297 0.317 0.336 0.292 ...
 $ net_change        : num [1:69] 0.164 0.157 0.156 -0.34 0.162 0.169 0.162 0.163 0.041 0.138 ...
 $ frac_rising       : num [1:69] 1 1 1 0.3 1 1 1 1 0.35 1 ...
 $ max_step          : num [1:69] 0.017 0.019 0.021 -0.042 0.016 ...
 $ typical_step      : num [1:69] 0.0055 0.005 0.005 0.019 0.006 ...
 $ step_excess       : num [1:69] -0.001 0.0015 0.002 -0.036 0.001 ...
 $ glitch_at_min     : num [1:69] 38 8 2 2 24 22 18 28 30 2 ...
 $ step_ratio        : num [1:69] 3.09 3.8 4.2 2.21 2.67 ...
 $ flag_decreasing   : logi [1:69] FALSE FALSE FALSE TRUE FALSE FALSE ...
 $ flag_over_range   : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:69] FALSE FALSE FALSE TRUE FALSE FALSE ...
 $ flag_high_baseline: logi [1:69] FALSE FALSE FALSE TRUE FALSE FALSE ...
 $ flag_bg_active    : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags           : int [1:69] 0 0 0 3 0 0 0 0 1 0 ...

5.2 Anomalous wells

anomalies <- well_diagnostics %>%
  filter(n_flags > 0) %>%
  mutate(flags = paste0(
    ifelse(flag_decreasing, "decreasing ", ""),
    ifelse(flag_over_range, "over-range ", ""),
    ifelse(flag_glitch,        "glitch ",        ""),
    ifelse(flag_bg_active,     "BG-active ",     ""),
    ifelse(flag_high_baseline, "high-baseline ", "")
  ) %>% trimws()) %>%
  arrange(well_type, plate_row, plate_col)

kable(anomalies %>%
        select(well, well_type, sample_id, std_nmol, od_first, od_last,
               net_change, frac_rising, od_max, flags),
      digits = c(0, 0, 0, 0, 3, 3, 3, 2, 3, 0),
      col.names = c("Well", "Type", "Sample", "STD nmol", "A412 t0", "A412 t40",
                    "Net dA412", "Frac rising", "Max A412", "Flags"),
      caption = "Wells failing one or more trace-shape checks")

write.csv(anomalies %>% select(-flag_decreasing, -flag_over_range, -flag_glitch,
                               -flag_bg_active, -flag_high_baseline, -n_flags),
          file.path(output_dir, "anomalous_wells.csv"), row.names = FALSE)

cat("\n--- anomalies: wells failing >=1 trace-shape check ---\n\n")
str(anomalies)
Well Type Sample STD nmol A412 t0 A412 t40 Net dA412 Frac rising Max A412 Flags
C1 background F05_01_ambient NA 0.057 0.057 0.000 0.35 0.064 decreasing
C2 background F05_01_ambient NA 0.405 0.691 0.286 0.95 0.691 BG-active
C5 background F05_02_ambient NA 0.812 0.453 -0.359 0.05 0.813 decreasing BG-active
C7 background F05_03_ambient NA 0.748 0.342 -0.406 0.00 0.748 decreasing BG-active
C12 background F05_04_ambient NA 0.566 0.260 -0.306 0.00 0.566 decreasing BG-active
D1 background F05_05_ambient NA 0.237 0.079 -0.158 0.00 0.237 decreasing BG-active
D2 background F05_05_ambient NA 0.097 0.060 -0.037 0.05 0.097 decreasing
D4 background F05_06_ambient NA 0.301 0.077 -0.224 0.00 0.301 decreasing glitch BG-active
D5 background F05_06_ambient NA 0.817 0.409 -0.408 0.00 0.817 decreasing BG-active
A6 sample F05_02_ambient NA 0.287 0.328 0.041 0.35 0.336 glitch
A12 sample F05_04_ambient NA 0.671 0.331 -0.340 0.30 0.671 decreasing glitch high-baseline
B1 sample F05_05_ambient NA 0.650 0.968 0.318 0.65 1.080 glitch high-baseline
B2 sample F05_05_ambient NA 0.669 0.925 0.256 0.70 0.973 glitch high-baseline
B3 sample F05_05_ambient NA 0.929 0.667 -0.262 0.15 0.954 decreasing high-baseline
B5 sample F05_06_ambient NA 0.841 0.696 -0.145 0.10 0.878 decreasing glitch high-baseline
E1 standard NA 0 0.100 0.100 0.000 0.00 0.100 decreasing
E2 standard NA 0 0.871 0.418 -0.453 0.05 0.871 decreasing
E4 standard NA 8 0.406 0.372 -0.034 0.00 0.406 decreasing
E5 standard NA 8 0.421 0.384 -0.037 0.00 0.421 decreasing
E6 standard NA 8 1.222 0.483 -0.739 0.00 1.222 decreasing glitch
E7 standard NA 16 1.349 1.099 -0.250 0.05 1.349 decreasing glitch
E8 standard NA 16 0.777 0.666 -0.111 0.00 0.777 decreasing
E9 standard NA 16 1.167 0.841 -0.326 0.00 1.167 decreasing
E10 standard NA 24 1.791 1.473 -0.318 0.15 1.810 decreasing over-range glitch
E11 standard NA 24 1.425 1.108 -0.317 0.15 1.454 decreasing glitch
E12 standard NA 24 0.890 0.841 -0.049 0.00 0.890 decreasing
F1 standard NA 32 2.052 1.572 -0.480 0.05 2.052 decreasing over-range glitch
F2 standard NA 32 2.135 1.542 -0.593 0.00 2.135 decreasing over-range
F3 standard NA 32 1.695 1.329 -0.366 0.00 1.695 decreasing over-range
F4 standard NA 40 2.189 1.737 -0.452 0.00 2.189 decreasing over-range
F5 standard NA 40 1.929 1.614 -0.315 0.00 1.929 decreasing over-range
F6 standard NA 40 2.101 1.738 -0.363 0.00 2.101 decreasing over-range

Wells failing one or more trace-shape checks

--- anomalies: wells failing >=1 trace-shape check ---

tibble [32 × 25] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:32] "C1" "C2" "C5" "C7" ...
 $ plate_row         : chr [1:32] "C" "C" "C" "C" ...
 $ plate_col         : int [1:32] 1 2 5 7 12 1 2 4 5 6 ...
 $ well_type         : chr [1:32] "background" "background" "background" "background" ...
 $ sample_id         : chr [1:32] "F05_01_ambient" "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" ...
 $ label             : chr [1:32] "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_02_ambient_BG-citrate_synthase-17.5-df.0" "F05_03_ambient_BG-citrate_synthase-22.3-df.0" ...
 $ std_nmol          : num [1:32] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:32] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:32] 0.057 0.405 0.812 0.748 0.566 0.237 0.097 0.301 0.817 0.287 ...
 $ od_last           : num [1:32] 0.057 0.691 0.453 0.342 0.26 0.079 0.06 0.077 0.409 0.328 ...
 $ od_max            : num [1:32] 0.064 0.691 0.813 0.748 0.566 0.237 0.097 0.301 0.817 0.336 ...
 $ net_change        : num [1:32] 0 0.286 -0.359 -0.406 -0.306 -0.158 -0.037 -0.224 -0.408 0.041 ...
 $ frac_rising       : num [1:32] 0.35 0.95 0.05 0 0 0 0.05 0 0 0.35 ...
 $ max_step          : num [1:32] -0.004 0.028 -0.036 -0.043 -0.023 ...
 $ typical_step      : num [1:32] 0 0.016 0.021 0.0155 0.015 ...
 $ step_excess       : num [1:32] 0.004 -0.02 0.018 -0.005 -0.0075 ...
 $ glitch_at_min     : num [1:32] 40 4 2 14 24 6 16 2 34 30 ...
 $ step_ratio        : num [1:32] 4 1.75 1.71 2.77 1.53 ...
 $ flag_decreasing   : logi [1:32] TRUE FALSE TRUE TRUE TRUE TRUE ...
 $ flag_over_range   : logi [1:32] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:32] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:32] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:32] FALSE TRUE TRUE TRUE TRUE TRUE ...
 $ n_flags           : int [1:32] 1 1 2 2 2 2 1 3 2 1 ...
 $ flags             : chr [1:32] "decreasing" "BG-active" "decreasing BG-active" "decreasing BG-active" ...

5.3 Background control wells behaving as reactions

This is the single largest anomaly on the plate and is called out separately. Background wells receive Background Control Mix, which contains no CS Substrate Mix, so they cannot sustain a CS reaction and must read flat.

bg_summary <- well_diagnostics %>%
  filter(well_type == "background") %>%
  mutate(behaviour = ifelse(flag_bg_active, "ANOMALOUS", "flat (expected)")) %>%
  arrange(sample_id, well)

cat("Background wells by behaviour:\n")
print(bg_summary %>% count(behaviour) %>% as.data.frame(), row.names = FALSE)

cat("\nBackground triplicates, per sample:\n")
print(bg_summary %>%
        group_by(sample_id) %>%
        summarise(n_flat = sum(!flag_bg_active), n_anomalous = sum(flag_bg_active),
                  anomalous_wells = paste(well[flag_bg_active], collapse = ", "),
                  .groups = "drop") %>% as.data.frame(), row.names = FALSE)

kable(bg_summary %>% select(well, sample_id, od_first, od_last, net_change, behaviour),
      digits = c(0, 0, 3, 3, 3, 0),
      col.names = c("Well", "Sample", "A412 t0", "A412 t40", "Net dA412", "Behaviour"),
      caption = "Background control wells: starting A412 and total drift")

cat("\n--- bg_summary: background well behaviour classification ---\n\n")
str(bg_summary)
Background wells by behaviour:
       behaviour  n
       ANOMALOUS  7
 flat (expected) 17

Background triplicates, per sample:
      sample_id n_flat n_anomalous anomalous_wells
 F05_01_ambient      2           1              C2
 F05_02_ambient      2           1              C5
 F05_03_ambient      2           1              C7
 F05_04_ambient      2           1             C12
 F05_05_ambient      2           1              D1
 F05_06_ambient      1           2          D4, D5
 F05_07_ambient      3           0                
 F05_08_ambient      3           0
Well Sample A412 t0 A412 t40 Net dA412 Behaviour
C1 F05_01_ambient 0.057 0.057 0.000 flat (expected)
C2 F05_01_ambient 0.405 0.691 0.286 ANOMALOUS
C3 F05_01_ambient 0.051 0.060 0.009 flat (expected)
C4 F05_02_ambient 0.055 0.059 0.004 flat (expected)
C5 F05_02_ambient 0.812 0.453 -0.359 ANOMALOUS
C6 F05_02_ambient 0.051 0.054 0.003 flat (expected)
C7 F05_03_ambient 0.748 0.342 -0.406 ANOMALOUS
C8 F05_03_ambient 0.059 0.065 0.006 flat (expected)
C9 F05_03_ambient 0.050 0.057 0.007 flat (expected)
C10 F05_04_ambient 0.056 0.061 0.005 flat (expected)
C11 F05_04_ambient 0.050 0.057 0.007 flat (expected)
C12 F05_04_ambient 0.566 0.260 -0.306 ANOMALOUS
D1 F05_05_ambient 0.237 0.079 -0.158 ANOMALOUS
D2 F05_05_ambient 0.097 0.060 -0.037 flat (expected)
D3 F05_05_ambient 0.053 0.058 0.005 flat (expected)
D4 F05_06_ambient 0.301 0.077 -0.224 ANOMALOUS
D5 F05_06_ambient 0.817 0.409 -0.408 ANOMALOUS
D6 F05_06_ambient 0.063 0.070 0.007 flat (expected)
D7 F05_07_ambient 0.070 0.073 0.003 flat (expected)
D8 F05_07_ambient 0.061 0.067 0.006 flat (expected)
D9 F05_07_ambient 0.058 0.066 0.008 flat (expected)
D10 F05_08_ambient 0.061 0.068 0.007 flat (expected)
D11 F05_08_ambient 0.076 0.081 0.005 flat (expected)
D12 F05_08_ambient 0.050 0.054 0.004 flat (expected)

Background control wells: starting A412 and total drift

--- bg_summary: background well behaviour classification ---

tibble [24 × 25] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:24] "C1" "C2" "C3" "C4" ...
 $ plate_row         : chr [1:24] "C" "C" "C" "C" ...
 $ plate_col         : int [1:24] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_type         : chr [1:24] "background" "background" "background" "background" ...
 $ sample_id         : chr [1:24] "F05_01_ambient" "F05_01_ambient" "F05_01_ambient" "F05_02_ambient" ...
 $ label             : chr [1:24] "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_01_ambient_BG-citrate_synthase-10.1-df.0" "F05_02_ambient_BG-citrate_synthase-17.5-df.0" ...
 $ std_nmol          : num [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:24] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:24] 0.057 0.405 0.051 0.055 0.812 0.051 0.748 0.059 0.05 0.056 ...
 $ od_last           : num [1:24] 0.057 0.691 0.06 0.059 0.453 0.054 0.342 0.065 0.057 0.061 ...
 $ od_max            : num [1:24] 0.064 0.691 0.06 0.059 0.813 0.054 0.748 0.065 0.057 0.061 ...
 $ net_change        : num [1:24] 0 0.286 0.009 0.004 -0.359 0.003 -0.406 0.006 0.007 0.005 ...
 $ frac_rising       : num [1:24] 0.35 0.95 0.25 0.3 0.05 0.15 0 0.4 0.45 0.25 ...
 $ max_step          : num [1:24] -0.004 0.028 0.005 0.001 -0.036 ...
 $ typical_step      : num [1:24] 0 0.016 0 0 0.021 ...
 $ step_excess       : num [1:24] 0.004 -0.02 0.005 -0.002 0.018 ...
 $ glitch_at_min     : num [1:24] 40 4 36 22 2 20 14 2 14 2 ...
 $ step_ratio        : num [1:24] 4 1.75 5 1 1.71 ...
 $ flag_decreasing   : logi [1:24] TRUE FALSE FALSE FALSE TRUE FALSE ...
 $ flag_over_range   : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_high_baseline: logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_bg_active    : logi [1:24] FALSE TRUE FALSE FALSE TRUE FALSE ...
 $ n_flags           : int [1:24] 1 1 0 0 2 0 2 0 0 0 ...
 $ behaviour         : chr [1:24] "flat (expected)" "ANOMALOUS" "flat (expected)" "flat (expected)" ...

5.4 Starting absorbance as the master diagnostic

Every anomaly above — decreasing samples, active background wells, scattered standards — occurs in wells that started high. A412 at t = 0 reports TNB already present before the reaction began, so an elevated baseline means the well was contaminated or mis-dispensed. This matters for interpretation: a well that starts high and then declines is not measuring CS activity at all, and a rate extracted from it is an artifact of whatever is decaying.

baseline_diag       <- compute_baseline_diagnostics(well_diagnostics, assay_params)
baseline_check      <- baseline_diag$baseline_check
baseline_per_sample <- baseline_diag$baseline_per_sample

cat("Sample and background wells by starting A412:\n")
print(baseline_check %>%
        group_by(well_type, baseline) %>%
        summarise(n = n(), n_decreasing = sum(flag_decreasing), .groups = "drop") %>%
        as.data.frame(), row.names = FALSE)

# Does an elevated baseline track with a declining trace?
baseline_vs_direction <- baseline_check %>%
  summarise(
    elevated_and_decreasing = sum(baseline == "elevated" & flag_decreasing),
    elevated_total          = sum(baseline == "elevated"),
    normal_and_decreasing   = sum(baseline == "normal" & flag_decreasing),
    normal_total            = sum(baseline == "normal")
  )

cat("\nElevated baseline & decreasing:", baseline_vs_direction$elevated_and_decreasing,
    "/", baseline_vs_direction$elevated_total, "\n")
cat("Normal baseline & decreasing  :", baseline_vs_direction$normal_and_decreasing,
    "/", baseline_vs_direction$normal_total, "\n")

cat("\nSample wells with elevated baseline:\n")
print(baseline_check %>% filter(well_type == "sample", baseline == "elevated") %>%
        select(well, sample_id, od_first, net_change) %>%
        as.data.frame(), row.names = FALSE, digits = 3)

cat("\nElevated-baseline replicates per sample:\n")
print(as.data.frame(baseline_per_sample), row.names = FALSE, digits = 3)

cat("\n--- baseline_check: starting-A412 classification per sample/background well ---\n\n")
str(baseline_check)
cat("\n--- baseline_per_sample: count of compromised replicates per sample ---\n\n")
str(baseline_per_sample)
Sample and background wells by starting A412:
  well_type baseline  n n_decreasing
 background elevated  5            4
 background   normal 19            4
     sample elevated  5            3
     sample   normal 19            0

Elevated baseline & decreasing: 7 / 10 
Normal baseline & decreasing  : 4 / 38 

Sample wells with elevated baseline:
 well      sample_id od_first net_change
   B3 F05_05_ambient    0.929     -0.262
   B5 F05_06_ambient    0.841     -0.145
  A12 F05_04_ambient    0.671     -0.340
   B2 F05_05_ambient    0.669      0.256
   B1 F05_05_ambient    0.650      0.318

Elevated-baseline replicates per sample:
      sample_id n_elevated n median_baseline
 F05_05_ambient          3 3           0.669
 F05_04_ambient          1 3           0.136
 F05_06_ambient          1 3           0.322
 F05_01_ambient          0 3           0.126
 F05_02_ambient          0 3           0.154
 F05_03_ambient          0 3           0.154
 F05_07_ambient          0 3           0.149
 F05_08_ambient          0 3           0.193

--- baseline_check: starting-A412 classification per sample/background well ---

tibble [48 × 25] (S3: tbl_df/tbl/data.frame)
 $ well              : chr [1:48] "B3" "B5" "D5" "C5" ...
 $ plate_row         : chr [1:48] "B" "B" "D" "C" ...
 $ plate_col         : int [1:48] 3 5 5 5 7 12 2 1 12 2 ...
 $ well_type         : chr [1:48] "sample" "sample" "background" "background" ...
 $ sample_id         : chr [1:48] "F05_05_ambient" "F05_06_ambient" "F05_06_ambient" "F05_02_ambient" ...
 $ label             : chr [1:48] "F05_05_ambient-citrate_synthase-11.5-df.0" "F05_06_ambient-citrate_synthase-30.2-df.0" "F05_06_ambient_BG-citrate_synthase-30.2-df.0" "F05_02_ambient_BG-citrate_synthase-17.5-df.0" ...
 $ std_nmol          : num [1:48] NA NA NA NA NA NA NA NA NA NA ...
 $ source            : chr [1:48] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ od_first          : num [1:48] 0.929 0.841 0.817 0.812 0.748 0.671 0.669 0.65 0.566 0.405 ...
 $ od_last           : num [1:48] 0.667 0.696 0.409 0.453 0.342 0.331 0.925 0.968 0.26 0.691 ...
 $ od_max            : num [1:48] 0.954 0.878 0.817 0.813 0.748 0.671 0.973 1.08 0.566 0.691 ...
 $ net_change        : num [1:48] -0.262 -0.145 -0.408 -0.359 -0.406 -0.34 0.256 0.318 -0.306 0.286 ...
 $ frac_rising       : num [1:48] 0.15 0.1 0 0.05 0 0.3 0.7 0.65 0 0.95 ...
 $ max_step          : num [1:48] -0.033 0.029 -0.028 -0.036 -0.043 ...
 $ typical_step      : num [1:48] 0.0145 0.01 0.0205 0.021 0.0155 ...
 $ step_excess       : num [1:48] -0.019 0.031 0.004 0.018 -0.005 ...
 $ glitch_at_min     : num [1:48] 20 2 34 2 14 2 4 4 24 4 ...
 $ step_ratio        : num [1:48] 2.28 2.9 1.37 1.71 2.77 ...
 $ flag_decreasing   : logi [1:48] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ flag_over_range   : logi [1:48] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch       : logi [1:48] FALSE TRUE FALSE FALSE FALSE TRUE ...
 $ flag_high_baseline: logi [1:48] TRUE TRUE FALSE FALSE FALSE TRUE ...
 $ flag_bg_active    : logi [1:48] FALSE FALSE TRUE TRUE TRUE FALSE ...
 $ n_flags           : int [1:48] 2 3 2 2 2 3 2 2 2 1 ...
 $ baseline          : chr [1:48] "elevated" "elevated" "elevated" "elevated" ...

--- baseline_per_sample: count of compromised replicates per sample ---

tibble [8 × 4] (S3: tbl_df/tbl/data.frame)
 $ sample_id      : chr [1:8] "F05_05_ambient" "F05_04_ambient" "F05_06_ambient" "F05_01_ambient" ...
 $ n_elevated     : int [1:8] 3 1 1 0 0 0 0 0
 $ n              : int [1:8] 3 3 3 3 3 3 3 3
 $ median_baseline: num [1:8] 0.669 0.136 0.322 0.126 0.154 0.154 0.149 0.193
baseline_plot <- ggplot(baseline_check,
                        aes(x = od_first, y = net_change, colour = well_type)) +
  geom_hline(yintercept = 0, linetype = "dashed", colour = "grey50") +
  geom_vline(xintercept = assay_params$sample_baseline_od,
             linetype = "dotted", colour = "grey40") +
  geom_point(size = 3, alpha = 0.85) +
  geom_text(aes(label = well), vjust = -0.9, size = 2.6, show.legend = FALSE) +
  scale_colour_manual(values = c(sample = "#08519c", background = "#a63603"),
                      name = "Well type") +
  labs(title = "Starting absorbance predicts trace direction",
       subtitle = paste0("Clean sample wells cluster tightly at low A412 with a positive net change.\n",
                         "Wells starting above A412 = ", assay_params$sample_baseline_od,
                         " (dotted) scatter and mostly decline: TNB was\n",
                         "already present before the read began, so their traces are not enzymology."),
       x = "A412 at t = 0", y = "Net dA412 over 40 min") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "baseline_vs_direction.png"), baseline_plot,
       width = 9, height = 6, dpi = 300)

cat("--- baseline_plot: ggplot object structure ---\n\n")
summary(baseline_plot)

baseline_plot

--- baseline_plot: ggplot object structure ---

data: well, plate_row, plate_col, well_type, sample_id, label,
  std_nmol, source, od_first, od_last, od_max, net_change, frac_rising,
  max_step, typical_step, step_excess, glitch_at_min, step_ratio,
  flag_decreasing, flag_over_range, flag_glitch, flag_high_baseline,
  flag_bg_active, n_flags, baseline [48x25]
mapping:  x = ~od_first, y = ~net_change, colour = ~well_type
scales:   colour 
faceting:  <empty> 
-----------------------------------
mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: xintercept = ~xintercept 
geom_vline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: label = ~well 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge

6 GSH STANDARD CURVE

The GSH standard is a stoichiometric endpoint, not an enzymatic reaction: free thiol reduces DTNB immediately on mixing. The standard curve is therefore read from a single timepoint, and t = 0 is used because TNB2- degrades over the run (quantified below).

6.1 Fit the standard curve

Three fits are compared, so the effect of the outlier wells on the calibration is explicit rather than buried:

  • all wells — every standard replicate, no exclusions
  • concentration means — the conventional fit, on per-concentration means
  • outlier-excluded — replicate-level fit after dropping flagged wells

The outlier-excluded fit is used for activity calculation. Standards are flagged by deviation from their own triplicate median, which is not dragged by a single contaminated well the way a mean is.

std_curve <- fit_standard_curve(plate_long, assay_params)

standards_t0       <- std_curve$standards_t0
standard_drift     <- std_curve$standard_drift
standard_summary   <- std_curve$standard_summary
standards_flagged  <- std_curve$standards_flagged
fit_all_wells      <- std_curve$fit_all_wells
fit_comparison     <- std_curve$fit_comparison
std_slope          <- std_curve$std_slope
std_intercept      <- std_curve$std_intercept
std_r2             <- std_curve$std_r2
std_nmol_max       <- std_curve$std_nmol_max

cat("Standard wells:", nrow(standards_t0), "at", dplyr::n_distinct(standards_t0$std_nmol),
    "concentrations\n")
cat("Standard drift over 40 min (dA412): median",
    sprintf("%.3f", median(standard_drift$drift)),
    "| range", paste(sprintf("%.3f", range(standard_drift$drift)), collapse = " to "), "\n")
cat("Standard wells that DECREASED:", sum(standard_drift$drift < 0), "/", nrow(standard_drift), "\n")

kable(standard_drift, digits = c(0, 0, 3, 3, 3),
      col.names = c("Well", "GSH (nmol)", "A412 t0", "A412 t40", "Drift"),
      caption = "GSH standard wells lose signal over the kinetic run (TNB instability)")

kable(standard_summary %>% select(std_nmol, n, mean_od, median_od, sd_od, se_od,
                                  cv_pct, net_od, od_per_nmol),
      digits = c(0, 0, 3, 3, 3, 3, 1, 3, 4),
      col.names = c("GSH (nmol)", "n", "Mean A412", "Median A412", "SD", "SEM",
                    "CV (%)", "Net A412", "A412/nmol"),
      caption = "GSH standard curve summary statistics (t = 0)")

cat("Standard concentrations exceeding the", assay_params$cv_threshold_pct,
    "% CV threshold:",
    paste(standard_summary$std_nmol[standard_summary$cv_pct > assay_params$cv_threshold_pct],
          collapse = ", "), "\n")

kable(standards_flagged %>% select(well, std_nmol, od, triplicate_median,
                                    deviation, is_outlier),
      digits = c(0, 0, 3, 3, 3, 0),
      col.names = c("Well", "GSH (nmol)", "A412", "Triplicate median",
                    "Deviation", "Outlier"),
      caption = paste0("Standard wells ranked by deviation from triplicate median (threshold ",
                       assay_params$std_outlier_od, " A412)"))

cat("Standard wells flagged as outliers:", sum(standards_flagged$is_outlier),
    "/", nrow(standards_flagged), "\n")

kable(fit_comparison, digits = c(0, 0, 5, 4, 4),
      col.names = c("Fit", "n", "Slope (A412/nmol)", "Intercept", "R-squared"),
      caption = "GSH standard curve: three fitting strategies compared")

cat("\nCalibration used downstream (outlier-excluded fit):\n")
cat("  A412 = ", sprintf("%.5f", std_slope), " * nmol + ",
    sprintf("%.4f", std_intercept), "   R^2 = ", sprintf("%.4f", std_r2), "\n", sep = "")
cat("  1 nmol thiol =", sprintf("%.5f", std_slope), "A412 units\n")

cat("\n--- standards_flagged: standard wells with median-deviation outlier flags ---\n\n")
str(standards_flagged)
cat("\n--- fit_comparison: side-by-side standard curve fit statistics ---\n\n")
str(fit_comparison)
Standard wells: 18 at 6 concentrations
Standard drift over 40 min (dA412): median -0.317 | range -0.739 to 0.004 
Standard wells that DECREASED: 16 / 18
Well GSH (nmol) A412 t0 A412 t40 Drift
E1 0 0.100 0.100 0.000
E2 0 0.871 0.418 -0.453
E3 0 0.095 0.099 0.004
E4 8 0.406 0.372 -0.034
E5 8 0.421 0.384 -0.037
E6 8 1.222 0.483 -0.739
E7 16 1.349 1.099 -0.250
E8 16 0.777 0.666 -0.111
E9 16 1.167 0.841 -0.326
E10 24 1.791 1.473 -0.318
E11 24 1.425 1.108 -0.317
E12 24 0.890 0.841 -0.049
F1 32 2.052 1.572 -0.480
F2 32 2.135 1.542 -0.593
F3 32 1.695 1.329 -0.366
F4 40 2.189 1.737 -0.452
F5 40 1.929 1.614 -0.315
F6 40 2.101 1.738 -0.363

GSH standard wells lose signal over the kinetic run (TNB instability)

GSH (nmol) n Mean A412 Median A412 SD SEM CV (%) Net A412 A412/nmol
0 3 0.355 0.100 0.447 0.258 125.7 0.000 NA
8 3 0.683 0.421 0.467 0.270 68.4 0.328 0.0410
16 3 1.098 1.167 0.292 0.169 26.6 0.742 0.0464
24 3 1.369 1.425 0.453 0.262 33.1 1.013 0.0422
32 3 1.961 2.052 0.234 0.135 11.9 1.605 0.0502
40 3 2.073 2.101 0.132 0.076 6.4 1.718 0.0429

GSH standard curve summary statistics (t = 0)

Standard concentrations exceeding the 15 % CV threshold: 0, 8, 16, 24
Well GSH (nmol) A412 Triplicate median Deviation Outlier
E6 8 1.222 0.421 0.801 TRUE
E2 0 0.871 0.100 0.771 TRUE
E12 24 0.890 1.425 -0.535 TRUE
E8 16 0.777 1.167 -0.390 TRUE
E10 24 1.791 1.425 0.366 TRUE
F3 32 1.695 2.052 -0.357 TRUE
E7 16 1.349 1.167 0.182 TRUE
F5 40 1.929 2.101 -0.172 TRUE
F4 40 2.189 2.101 0.088 FALSE
F2 32 2.135 2.052 0.083 FALSE
E4 8 0.406 0.421 -0.015 FALSE
E3 0 0.095 0.100 -0.005 FALSE
E1 0 0.100 0.100 0.000 FALSE
E5 8 0.421 0.421 0.000 FALSE
E9 16 1.167 1.167 0.000 FALSE
E11 24 1.425 1.425 0.000 FALSE
F1 32 2.052 2.052 0.000 FALSE
F6 40 2.101 2.101 0.000 FALSE

Standard wells ranked by deviation from triplicate median (threshold 0.15 A412)

Standard wells flagged as outliers: 8 / 18
Fit n Slope (A412/nmol) Intercept R-squared
std_nmol…1 all wells 18 0.04533 0.3498 0.8042
std_nmol…2 concentration means 6 0.04533 0.3498 0.9829
std_nmol…3 outlier-excluded 10 0.05582 0.0927 0.9678

GSH standard curve: three fitting strategies compared

Calibration used downstream (outlier-excluded fit):
  A412 = 0.05582 * nmol + 0.0927   R^2 = 0.9678
  1 nmol thiol = 0.05582 A412 units

--- standards_flagged: standard wells with median-deviation outlier flags ---

tibble [18 × 7] (S3: tbl_df/tbl/data.frame)
 $ well             : chr [1:18] "E6" "E2" "E12" "E8" ...
 $ std_nmol         : num [1:18] 8 0 24 16 24 32 16 40 40 32 ...
 $ od               : num [1:18] 1.222 0.871 0.89 0.777 1.791 ...
 $ source           : chr [1:18] "absorbance_csv" "absorbance_csv" "absorbance_csv" "absorbance_csv" ...
 $ triplicate_median: num [1:18] 0.421 0.1 1.425 1.167 1.425 ...
 $ deviation        : num [1:18] 0.801 0.771 -0.535 -0.39 0.366 ...
 $ is_outlier       : logi [1:18] TRUE TRUE TRUE TRUE TRUE TRUE ...

--- fit_comparison: side-by-side standard curve fit statistics ---

'data.frame':   3 obs. of  5 variables:
 $ fit      : chr  "all wells" "concentration means" "outlier-excluded"
 $ n        : int  18 6 10
 $ slope    : num  0.0453 0.0453 0.0558
 $ intercept: num  0.3498 0.3498 0.0927
 $ r_squared: num  0.804 0.983 0.968

6.2 Plot the standard curve

std_curve_plot <- ggplot(standards_flagged, aes(x = std_nmol, y = od)) +
  geom_abline(aes(slope = coef(fit_all_wells)[2], intercept = coef(fit_all_wells)[1],
                  linetype = "all wells"), colour = "grey55") +
  geom_abline(aes(slope = std_slope, intercept = std_intercept,
                  linetype = "outlier-excluded"), colour = "#238b45", linewidth = 0.8) +
  geom_point(aes(shape = is_outlier, colour = is_outlier), size = 3) +
  geom_hline(yintercept = assay_params$od_linear_max, linetype = "dotted", colour = "grey40") +
  scale_shape_manual(values = c(`FALSE` = 16, `TRUE` = 4),
                     labels = c(`FALSE` = "retained", `TRUE` = "outlier"), name = "Well") +
  scale_colour_manual(values = c(`FALSE` = "#238b45", `TRUE` = "#cb181d"),
                      labels = c(`FALSE` = "retained", `TRUE` = "outlier"), name = "Well") +
  scale_linetype_manual(values = c(`all wells` = "dashed", `outlier-excluded` = "solid"),
                        name = "Fit") +
  labs(title = "GSH standard curve, A412 at t = 0",
       subtitle = paste0("Dotted line = photometric linearity ceiling (A412 = ",
                         assay_params$od_linear_max,
                         "); the 32 and 40 nmol standards sit above it"),
       x = "GSH (nmol/well)", y = "A412") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"))

ggsave(file.path(output_dir, "gsh_standard_curve.png"), std_curve_plot,
       width = 9, height = 6, dpi = 300)

cat("--- std_curve_plot: ggplot object structure ---\n\n")
summary(std_curve_plot)

std_curve_plot

--- std_curve_plot: ggplot object structure ---

data: well, std_nmol, od, source, triplicate_median, deviation,
  is_outlier [18x7]
mapping:  x = ~std_nmol, y = ~od
scales:   shape, colour, linetype 
faceting:  <empty> 
-----------------------------------
mapping: slope = ~coef(fit_all_wells)[2], intercept = ~coef(fit_all_wells)[1], linetype = all wells 
geom_abline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: slope = ~std_slope, intercept = ~std_intercept, linetype = outlier-excluded 
geom_abline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: shape = ~is_outlier, colour = ~is_outlier 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: yintercept = ~yintercept 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

7 RATE EXTRACTION

7.1 Sliding-window linear fit

Abcam §9.3 directs the analyst to pick two timepoints “in the linear range”. That is implemented here as a sliding window of 5 consecutive reads (8 min) scanned across each trace. Two candidate windows are recorded per well:

  • max increasing — the steepest positive slope, which is what CS activity means. This is the rate used for activity.
  • max absolute — the steepest slope of either sign. Gen5’s own Max V [412] column uses this convention, which is why Gen5 reports large negative “rates” for degrading wells.

Recording both makes the divergence between them an explicit QC signal.

well_rates <- compute_well_rates(plate_long, well_diagnostics, assay_params)

cat("Wells with a usable increasing rate:", sum(well_rates$rate_usable),
    "/", nrow(well_rates), "\n")
cat("Wells where the steepest window is NEGATIVE (Gen5 would report a negative Max V):",
    sum(well_rates$abs_window_is_negative), "\n")

cat("\n--- well_rates: per-well max-increasing and max-absolute slope windows ---\n\n")
str(well_rates)
Wells with a usable increasing rate: 36 / 69 
Wells where the steepest window is NEGATIVE (Gen5 would report a negative Max V): 27 

--- well_rates: per-well max-increasing and max-absolute slope windows ---

tibble [69 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:69] "A1" "A10" "A11" "A12" ...
 $ well_type             : chr [1:69] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:69] "F05_01_ambient" "F05_04_ambient" "F05_04_ambient" "F05_04_ambient" ...
 $ std_nmol              : num [1:69] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:69] 0 0 0 30 0 0 0 0 24 0 ...
 $ t_end                 : num [1:69] 8 8 8 38 8 8 8 8 32 8 ...
 $ slope_mOD_min         : num [1:69] 8.15 8.25 8.5 2.2 7.4 ...
 $ r2                    : num [1:69] 0.999 0.996 0.99 0.998 0.999 ...
 $ max_abs_slope_mOD_min : num [1:69] 8.15 8.25 8.5 -19.65 7.4 ...
 $ abs_window_is_negative: logi [1:69] FALSE FALSE FALSE TRUE FALSE FALSE ...
 $ net_change            : num [1:69] 0.164 0.157 0.156 -0.34 0.162 0.169 0.162 0.163 0.041 0.138 ...
 $ frac_rising           : num [1:69] 1 1 1 0.3 1 1 1 1 0.35 1 ...
 $ od_max                : num [1:69] 0.29 0.286 0.292 0.671 0.291 0.294 0.297 0.317 0.336 0.292 ...
 $ flag_decreasing       : logi [1:69] FALSE FALSE FALSE TRUE FALSE FALSE ...
 $ flag_over_range       : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:69] FALSE FALSE FALSE TRUE FALSE FALSE ...
 $ glitch_at_min         : num [1:69] 38 8 2 2 24 22 18 28 30 2 ...
 $ flag_bg_active        : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:69] 0 0 0 3 0 0 0 0 1 0 ...
 $ glitch_in_window      : logi [1:69] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:69] TRUE TRUE TRUE FALSE TRUE TRUE ...

7.2 Sample and background rates

sample_rates <- well_rates %>%
  filter(well_type == "sample") %>%
  arrange(sample_id, well)

background_rates <- well_rates %>%
  filter(well_type == "background") %>%
  arrange(sample_id, well)

kable(sample_rates %>% select(well, sample_id, t_start, t_end, slope_mOD_min, r2,
                              net_change, max_abs_slope_mOD_min,
                              glitch_in_window, rate_usable),
      digits = c(0, 0, 0, 0, 2, 3, 3, 2, 0, 0),
      col.names = c("Well", "Sample", "Window start (min)", "Window end (min)",
                    "Rate (mA412/min)", "R-squared", "Net dA412",
                    "Max |rate|", "Glitch in window", "Usable"),
      caption = "Sample wells: max-increasing rate window")

kable(background_rates %>% select(well, sample_id, slope_mOD_min, r2,
                                  net_change, flag_bg_active),
      digits = c(0, 0, 2, 3, 3, 0),
      col.names = c("Well", "Sample", "Rate (mA412/min)", "R-squared",
                    "Net dA412", "Anomalous"),
      caption = "Background control wells: rates should be ~0")

cat("\n--- sample_rates: rate windows for the sample wells ---\n\n")
str(sample_rates)
cat("\n--- background_rates: rate windows for the background wells ---\n\n")
str(background_rates)
Well Sample Window start (min) Window end (min) Rate (mA412/min) R-squared Net dA412 Max |rate| Glitch in window Usable
A1 F05_01_ambient 0 8 8.15 0.999 0.164 8.15 FALSE TRUE
A2 F05_01_ambient 0 8 7.40 0.999 0.162 7.40 FALSE TRUE
A3 F05_01_ambient 0 8 7.65 0.999 0.169 7.65 FALSE TRUE
A4 F05_02_ambient 0 8 8.55 0.997 0.162 8.55 FALSE TRUE
A5 F05_02_ambient 0 8 9.70 0.995 0.163 9.70 FALSE TRUE
A6 F05_02_ambient 24 32 5.15 0.695 0.041 5.15 TRUE FALSE
A7 F05_03_ambient 0 8 7.70 0.945 0.138 7.70 FALSE TRUE
A8 F05_03_ambient 0 8 8.50 0.950 0.152 8.50 FALSE TRUE
A9 F05_03_ambient 0 8 7.80 0.941 0.142 7.80 FALSE TRUE
A10 F05_04_ambient 0 8 8.25 0.996 0.157 8.25 FALSE TRUE
A11 F05_04_ambient 0 8 8.50 0.990 0.156 8.50 FALSE TRUE
A12 F05_04_ambient 30 38 2.20 0.998 -0.340 -19.65 FALSE FALSE
B1 F05_05_ambient 10 18 21.30 1.000 0.318 21.30 FALSE TRUE
B2 F05_05_ambient 10 18 17.20 0.994 0.256 17.20 FALSE TRUE
B3 F05_05_ambient 0 8 3.10 0.911 -0.262 -11.50 FALSE FALSE
B4 F05_06_ambient 0 8 6.15 0.824 0.050 6.15 FALSE TRUE
B5 F05_06_ambient 0 8 4.10 0.725 -0.145 -10.65 TRUE FALSE
B6 F05_06_ambient 0 8 8.75 0.955 0.150 8.75 FALSE TRUE
B7 F05_07_ambient 0 8 9.55 0.971 0.152 9.55 FALSE TRUE
B8 F05_07_ambient 0 8 8.40 0.975 0.148 8.40 FALSE TRUE
B9 F05_07_ambient 0 8 8.80 0.974 0.148 8.80 FALSE TRUE
B10 F05_08_ambient 0 8 7.65 0.854 0.120 7.65 FALSE TRUE
B11 F05_08_ambient 0 8 6.80 0.878 0.123 6.80 FALSE TRUE
B12 F05_08_ambient 0 8 6.60 0.837 0.119 6.60 FALSE TRUE

Sample wells: max-increasing rate window

Well Sample Rate (mA412/min) R-squared Net dA412 Anomalous
C1 F05_01_ambient 0.25 0.893 0.000 FALSE
C2 F05_01_ambient 11.55 0.989 0.286 TRUE
C3 F05_01_ambient 0.85 0.830 0.009 FALSE
C4 F05_02_ambient 0.15 0.750 0.004 FALSE
C5 F05_02_ambient -3.35 0.885 -0.359 TRUE
C6 F05_02_ambient 0.80 0.653 0.003 FALSE
C7 F05_03_ambient -5.25 1.000 -0.406 TRUE
C8 F05_03_ambient 0.50 1.000 0.006 FALSE
C9 F05_03_ambient 0.40 0.941 0.007 FALSE
C10 F05_04_ambient 0.20 0.800 0.005 FALSE
C11 F05_04_ambient 0.25 0.893 0.007 FALSE
C12 F05_04_ambient -5.15 1.000 -0.306 TRUE
D1 F05_05_ambient -1.45 0.987 -0.158 TRUE
D2 F05_05_ambient 0.05 0.083 -0.037 FALSE
D3 F05_05_ambient 0.25 0.893 0.005 FALSE
D4 F05_06_ambient -1.10 0.992 -0.224 TRUE
D5 F05_06_ambient -7.75 1.000 -0.408 TRUE
D6 F05_06_ambient 0.50 1.000 0.007 FALSE
D7 F05_07_ambient 0.15 0.750 0.003 FALSE
D8 F05_07_ambient 0.25 0.893 0.006 FALSE
D9 F05_07_ambient 0.50 1.000 0.008 FALSE
D10 F05_08_ambient 0.25 0.893 0.007 FALSE
D11 F05_08_ambient 0.25 0.893 0.005 FALSE
D12 F05_08_ambient 0.25 0.893 0.004 FALSE

Background control wells: rates should be ~0

--- sample_rates: rate windows for the sample wells ---

tibble [24 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:24] "A1" "A2" "A3" "A4" ...
 $ well_type             : chr [1:24] "sample" "sample" "sample" "sample" ...
 $ sample_id             : chr [1:24] "F05_01_ambient" "F05_01_ambient" "F05_01_ambient" "F05_02_ambient" ...
 $ std_nmol              : num [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:24] 0 0 0 0 0 24 0 0 0 0 ...
 $ t_end                 : num [1:24] 8 8 8 8 8 32 8 8 8 8 ...
 $ slope_mOD_min         : num [1:24] 8.15 7.4 7.65 8.55 9.7 ...
 $ r2                    : num [1:24] 0.999 0.999 0.999 0.997 0.995 ...
 $ max_abs_slope_mOD_min : num [1:24] 8.15 7.4 7.65 8.55 9.7 ...
 $ abs_window_is_negative: logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ net_change            : num [1:24] 0.164 0.162 0.169 0.162 0.163 0.041 0.138 0.152 0.142 0.157 ...
 $ frac_rising           : num [1:24] 1 1 1 1 1 0.35 1 1 1 1 ...
 $ od_max                : num [1:24] 0.29 0.291 0.294 0.297 0.317 0.336 0.292 0.307 0.296 0.286 ...
 $ flag_decreasing       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_over_range       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:24] FALSE FALSE FALSE FALSE FALSE TRUE ...
 $ glitch_at_min         : num [1:24] 38 24 22 18 28 30 2 2 2 8 ...
 $ flag_bg_active        : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_flags               : int [1:24] 0 0 0 0 0 1 0 0 0 0 ...
 $ glitch_in_window      : logi [1:24] FALSE FALSE FALSE FALSE FALSE TRUE ...
 $ rate_usable           : logi [1:24] TRUE TRUE TRUE TRUE TRUE FALSE ...

--- background_rates: rate windows for the background wells ---

tibble [24 × 21] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:24] "C1" "C2" "C3" "C4" ...
 $ well_type             : chr [1:24] "background" "background" "background" "background" ...
 $ sample_id             : chr [1:24] "F05_01_ambient" "F05_01_ambient" "F05_01_ambient" "F05_02_ambient" ...
 $ std_nmol              : num [1:24] NA NA NA NA NA NA NA NA NA NA ...
 $ t_start               : num [1:24] 16 24 30 0 0 16 32 10 24 10 ...
 $ t_end                 : num [1:24] 24 32 38 8 8 24 40 18 32 18 ...
 $ slope_mOD_min         : num [1:24] 0.25 11.55 0.85 0.15 -3.35 ...
 $ r2                    : num [1:24] 0.893 0.989 0.83 0.75 0.885 ...
 $ max_abs_slope_mOD_min : num [1:24] -0.95 11.55 0.85 0.15 -13.5 ...
 $ abs_window_is_negative: logi [1:24] TRUE FALSE FALSE FALSE TRUE FALSE ...
 $ net_change            : num [1:24] 0 0.286 0.009 0.004 -0.359 0.003 -0.406 0.006 0.007 0.005 ...
 $ frac_rising           : num [1:24] 0.35 0.95 0.25 0.3 0.05 0.15 0 0.4 0.45 0.25 ...
 $ od_max                : num [1:24] 0.064 0.691 0.06 0.059 0.813 0.054 0.748 0.065 0.057 0.061 ...
 $ flag_decreasing       : logi [1:24] TRUE FALSE FALSE FALSE TRUE FALSE ...
 $ flag_over_range       : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ flag_glitch           : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ glitch_at_min         : num [1:24] 40 4 36 22 2 20 14 2 14 2 ...
 $ flag_bg_active        : logi [1:24] FALSE TRUE FALSE FALSE TRUE FALSE ...
 $ n_flags               : int [1:24] 1 1 0 0 2 0 2 0 0 0 ...
 $ glitch_in_window      : logi [1:24] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ rate_usable           : logi [1:24] FALSE TRUE TRUE FALSE FALSE FALSE ...

7.3 Positive control

pos_control <- compute_positive_control(well_rates, std_slope, assay_params)

kable(pos_control %>% select(well, t_start, t_end, slope_mOD_min, r2, net_change),
      digits = c(0, 0, 0, 2, 4, 3),
      col.names = c("Well", "Window start (min)", "Window end (min)",
                    "Rate (mA412/min)", "R-squared", "Net dA412"),
      caption = "CS Positive Control: the plate's proof that the reaction chemistry worked")

cat("Positive control rate: mean", sprintf("%.2f", mean(pos_control$slope_mOD_min)),
    "mA412/min | CV", sprintf("%.1f", 100 * sd(pos_control$slope_mOD_min) /
                                     mean(pos_control$slope_mOD_min)), "%\n")
cat("All three replicates rising and linear:",
    all(pos_control$net_change > 0 & pos_control$r2 > 0.99), "\n")

cat("\n--- pos_control: positive control rate and R^2 per replicate ---\n\n")
str(pos_control)
Well Window start (min) Window end (min) Rate (mA412/min) R-squared Net dA412
F7 28 36 2.85 0.9991 0.093
F8 26 34 2.85 0.9991 0.099
F9 10 18 2.75 0.9990 0.104

CS Positive Control: the plate’s proof that the reaction chemistry worked

Positive control rate: mean 2.82 mA412/min | CV 2.0 %
All three replicates rising and linear: TRUE 

--- pos_control: positive control rate and R^2 per replicate ---

tibble [3 × 22] (S3: tbl_df/tbl/data.frame)
 $ well                  : chr [1:3] "F7" "F8" "F9"
 $ well_type             : chr [1:3] "positive_control" "positive_control" "positive_control"
 $ sample_id             : chr [1:3] NA NA NA
 $ std_nmol              : num [1:3] NA NA NA
 $ t_start               : num [1:3] 28 26 10
 $ t_end                 : num [1:3] 36 34 18
 $ slope_mOD_min         : num [1:3] 2.85 2.85 2.75
 $ r2                    : num [1:3] 0.999 0.999 0.999
 $ max_abs_slope_mOD_min : num [1:3] 2.85 2.85 2.75
 $ abs_window_is_negative: logi [1:3] FALSE FALSE FALSE
 $ net_change            : num [1:3] 0.093 0.099 0.104
 $ frac_rising           : num [1:3] 1 1 1
 $ od_max                : num [1:3] 0.322 0.335 0.32
 $ flag_decreasing       : logi [1:3] FALSE FALSE FALSE
 $ flag_over_range       : logi [1:3] FALSE FALSE FALSE
 $ flag_glitch           : logi [1:3] FALSE FALSE FALSE
 $ glitch_at_min         : num [1:3] 20 10 10
 $ flag_bg_active        : logi [1:3] FALSE FALSE FALSE
 $ n_flags               : int [1:3] 0 0 0
 $ glitch_in_window      : logi [1:3] FALSE FALSE FALSE
 $ rate_usable           : logi [1:3] TRUE TRUE TRUE
 $ activity_mU_uL        : num [1:3] 0.0255 0.0255 0.0246

8 BACKGROUND CORRECTION

Abcam §10.2 subtracts the background reading only “if sample background control reading is significant”. Here the background rate is the relevant quantity, and only the well-behaved (flat) background replicates are used to estimate it, since the anomalous ones do not measure background.

background_per_sample <- compute_background_correction(well_rates)

cat("Background rate from flat wells only (mA412/min):\n")
print(background_per_sample %>% as.data.frame(), row.names = FALSE, digits = 3)

cat("\nMedian flat-background rate:",
    sprintf("%.3f", median(background_per_sample$bg_rate_flat, na.rm = TRUE)),
    "mA412/min\n")
cat("Median sample rate:",
    sprintf("%.2f", median(sample_rates$slope_mOD_min)), "mA412/min\n")
cat("Background as % of sample signal:",
    sprintf("%.1f", 100 * median(background_per_sample$bg_rate_flat, na.rm = TRUE) /
                    median(sample_rates$slope_mOD_min)), "%\n")

cat("\n--- background_per_sample: per-sample background rate estimates ---\n\n")
str(background_per_sample)
Background rate from flat wells only (mA412/min):
      sample_id n_bg_total n_bg_flat bg_rate_flat bg_rate_all
 F05_01_ambient          3         2        0.550       4.217
 F05_02_ambient          3         2        0.475      -0.800
 F05_03_ambient          3         2        0.450      -1.450
 F05_04_ambient          3         2        0.225      -1.567
 F05_05_ambient          3         2        0.150      -0.383
 F05_06_ambient          3         1        0.500      -2.783
 F05_07_ambient          3         3        0.300       0.300
 F05_08_ambient          3         3        0.250       0.250

Median flat-background rate: 0.375 mA412/min
Median sample rate: 7.98 mA412/min
Background as % of sample signal: 4.7 %

--- background_per_sample: per-sample background rate estimates ---

tibble [8 × 5] (S3: tbl_df/tbl/data.frame)
 $ sample_id   : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_bg_total  : int [1:8] 3 3 3 3 3 3 3 3
 $ n_bg_flat   : int [1:8] 2 2 2 2 2 1 3 3
 $ bg_rate_flat: num [1:8] 0.55 0.475 0.45 0.225 0.15 ...
 $ bg_rate_all : num [1:8] 4.217 -0.8 -1.45 -1.567 -0.383 ...

8.1 Background significance test

Cattau et al. (2023; legacy version of this assay, Cattau_v1_master.Rmd) validated background subtraction once with an ANOVA and then applied that conclusion to every subsequent plate. The same formal test is run here – a Welch two-sample t-test of well-level rate (background vs. sample), plus an ANOVA on the OD trajectory over time – so the flat/active threshold classification used above is backed by an explicit significance test rather than the threshold alone.

bg_sample_rates <- well_rates %>%
  filter(well_type %in% c("sample", "background"))

bg_significance_test <- t.test(slope_mOD_min ~ well_type, data = bg_sample_rates)

cat("--- Welch two-sample t-test: well-level rate (background vs. sample) ---\n\n")
print(bg_significance_test)

bg_trend_anova <- anova(lm(od ~ well_type * time_min,
                            data = plate_long %>% filter(well_type %in% c("sample", "background"))))

cat("\n--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---\n\n")
print(bg_trend_anova)

cat("\nInterpretation: mean background-well rate is",
    sprintf("%.2f", unname(bg_significance_test$estimate["mean in group background"])),
    "mA412/min vs.", sprintf("%.2f", unname(bg_significance_test$estimate["mean in group sample"])),
    "mA412/min for sample wells (p =", format.pval(bg_significance_test$p.value, digits = 3),
    "). Background wells are statistically distinct from, and far slower\n",
    "than, reaction wells on this plate -- consistent with the per-well\n",
    "flat/active classification above and with the qualitative conclusion\n",
    "reached by Cattau et al. (2023) on the legacy assay version.\n")

cat("\n--- bg_significance_test: Welch t-test result object ---\n\n")
str(bg_significance_test)

cat("\n--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---\n\n")
str(bg_trend_anova)
--- Welch two-sample t-test: well-level rate (background vs. sample) ---


    Welch Two Sample t-test

data:  slope_mOD_min by well_type
t = -8.0493, df = 44.983, p-value = 2.931e-10
alternative hypothesis: true difference in means between group background and group sample is not equal to 0
95 percent confidence interval:
 -10.658164  -6.391836
sample estimates:
mean in group background     mean in group sample 
              -0.2770833                8.2479167 


--- ANOVA: OD trajectory by well_type x time_min (background vs sample wells) ---

Analysis of Variance Table

Response: od
                     Df Sum Sq Mean Sq  F value    Pr(>F)    
well_type             1  9.908  9.9081 205.3025 < 2.2e-16 ***
time_min              1  0.012  0.0117   0.2424  0.622564    
well_type:time_min    1  0.512  0.5118  10.6053  0.001165 ** 
Residuals          1004 48.454  0.0483                       
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Interpretation: mean background-well rate is -0.28 mA412/min vs. 8.25 mA412/min for sample wells (p = 2.93e-10 ). Background wells are statistically distinct from, and far slower
 than, reaction wells on this plate -- consistent with the per-well
 flat/active classification above and with the qualitative conclusion
 reached by Cattau et al. (2023) on the legacy assay version.

--- bg_significance_test: Welch t-test result object ---

List of 10
 $ statistic  : Named num -8.05
  ..- attr(*, "names")= chr "t"
 $ parameter  : Named num 45
  ..- attr(*, "names")= chr "df"
 $ p.value    : num 2.93e-10
 $ conf.int   : num [1:2] -10.66 -6.39
  ..- attr(*, "conf.level")= num 0.95
 $ estimate   : Named num [1:2] -0.277 8.248
  ..- attr(*, "names")= chr [1:2] "mean in group background" "mean in group sample"
 $ null.value : Named num 0
  ..- attr(*, "names")= chr "difference in means between group background and group sample"
 $ stderr     : num 1.06
 $ alternative: chr "two.sided"
 $ method     : chr "Welch Two Sample t-test"
 $ data.name  : chr "slope_mOD_min by well_type"
 - attr(*, "class")= chr "htest"

--- bg_trend_anova: ANOVA table for OD ~ well_type * time_min ---

Classes 'anova' and 'data.frame':   4 obs. of  5 variables:
 $ Df     : int  1 1 1 1004
 $ Sum Sq : num  9.9081 0.0117 0.5118 48.4542
 $ Mean Sq: num  9.9081 0.0117 0.5118 0.0483
 $ F value: num  205.302 0.242 10.605 NA
 $ Pr(>F) : num  1.67e-42 6.23e-01 1.17e-03 NA
 - attr(*, "heading")= chr [1:2] "Analysis of Variance Table\n" "Response: od"

9 TECHNICAL REPLICATE PRECISION

Coefficient of variation across the three technical replicates of each sample, computed on the extracted rate. Two versions are reported: all three replicates, and usable replicates only (rate window with R2 >= 0.8, rising trace, no read glitch). The comparison localizes the imprecision to specific wells. Any sample whose CV exceeds 15% on either basis is flagged below.

cv_summary <- compute_replicate_cv(well_rates, protein_by_sample$sample_id, assay_params)

kable(cv_summary %>% select(sample_id, n_all, mean_all, sd_all, cv_all,
                            n_usable, mean_usable, sd_usable, cv_usable,
                            excluded_wells),
      digits = c(0, 0, 2, 2, 1, 0, 2, 2, 1, 0),
      col.names = c("Sample", "n", "Mean rate", "SD", "CV (%)",
                    "n usable", "Mean rate", "SD", "CV (%)", "Excluded wells"),
      caption = paste0("Technical replicate CV of the extracted rate (mA412/min), ",
                       "all replicates vs. usable replicates only"))

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on ALL replicates:", sum(cv_summary$fails_cv_all), "/", nrow(cv_summary), "\n")
print(cv_summary %>% filter(fails_cv_all) %>%
        select(sample_id, n_all, mean_all, sd_all, cv_all) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

cat("\nSamples exceeding CV", paste0(assay_params$cv_threshold_pct, "%"),
    "on USABLE replicates:", sum(cv_summary$fails_cv_usable), "\n")
print(cv_summary %>% filter(fails_cv_usable) %>%
        select(sample_id, n_usable, mean_usable, sd_usable, cv_usable) %>%
        as.data.frame(), row.names = FALSE, digits = 4)

write.csv(cv_summary, file.path(output_dir, "technical_replicate_cv.csv"), row.names = FALSE)

cat("\n--- cv_summary: per-sample technical replicate CV, both variants ---\n\n")
str(cv_summary)
Sample n Mean rate SD CV (%) n usable Mean rate SD CV (%) Excluded wells
F05_01_ambient 3 7.73 0.38 4.9 3 7.73 0.38 4.9 -
F05_02_ambient 3 7.80 2.37 30.3 2 9.13 0.81 8.9 A6
F05_03_ambient 3 8.00 0.44 5.4 3 8.00 0.44 5.4 -
F05_04_ambient 3 6.32 3.57 56.5 2 8.38 0.18 2.1 A12
F05_05_ambient 3 13.87 9.55 68.8 2 19.25 2.90 15.1 B3
F05_06_ambient 3 6.33 2.33 36.8 2 7.45 1.84 24.7 B5
F05_07_ambient 3 8.92 0.58 6.5 3 8.92 0.58 6.5 -
F05_08_ambient 3 7.02 0.56 7.9 3 7.02 0.56 7.9 -

Technical replicate CV of the extracted rate (mA412/min), all replicates vs. usable replicates only

Samples exceeding CV 15% on ALL replicates: 4 / 8 
      sample_id n_all mean_all sd_all cv_all
 F05_02_ambient     3    7.800  2.366  30.33
 F05_04_ambient     3    6.317  3.567  56.47
 F05_05_ambient     3   13.867  9.547  68.85
 F05_06_ambient     3    6.333  2.330  36.80

Samples exceeding CV 15% on USABLE replicates: 2 
      sample_id n_usable mean_usable sd_usable cv_usable
 F05_05_ambient        2       19.25     2.899     15.06
 F05_06_ambient        2        7.45     1.838     24.68

--- cv_summary: per-sample technical replicate CV, both variants ---

'data.frame':   8 obs. of  12 variables:
 $ sample_id      : chr  "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_all          : int  3 3 3 3 3 3 3 3
 $ mean_all       : num  7.73 7.8 8 6.32 13.87 ...
 $ sd_all         : num  0.382 2.366 0.436 3.567 9.547 ...
 $ cv_all         : num  4.94 30.33 5.45 56.47 68.85 ...
 $ n_usable       : int  3 2 3 2 2 2 3 3
 $ mean_usable    : num  7.73 9.13 8 8.38 19.25 ...
 $ sd_usable      : num  0.382 0.813 0.436 0.177 2.899 ...
 $ cv_usable      : num  4.94 8.91 5.45 2.11 15.06 ...
 $ excluded_wells : Named chr  "-" "A6" "-" "A12" ...
  ..- attr(*, "names")= chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ fails_cv_all   : logi  FALSE TRUE FALSE TRUE TRUE TRUE ...
 $ fails_cv_usable: logi  FALSE FALSE FALSE FALSE TRUE TRUE ...

9.1 Plot replicate spread

cv_plot_data <- sample_rates %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id")

cv_plot <- ggplot(cv_plot_data, aes(x = sample_id, y = slope_mOD_min)) +
  geom_point(aes(colour = rate_usable, shape = rate_usable),
             size = 3, position = position_jitter(width = 0.12, height = 0)) +
  stat_summary(fun = mean, geom = "crossbar", width = 0.5,
               colour = "grey30", linewidth = 0.3) +
  geom_text(data = cv_summary, aes(x = sample_id, y = Inf,
                                   label = sprintf("CV %.0f%%", cv_all)),
            vjust = 1.6, size = 3,
            colour = ifelse(cv_summary$fails_cv_all, "#cb181d", "grey30")) +
  scale_colour_manual(values = c(`TRUE` = "#08519c", `FALSE` = "#cb181d"),
                      labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                     labels = c(`TRUE` = "usable", `FALSE` = "flagged"), name = "Replicate") +
  labs(title = "Technical replicate spread in extracted rate",
       subtitle = paste0("Crossbar = mean of all three replicates; CV computed on all three. ",
                         "Threshold ", assay_params$cv_threshold_pct, "%."),
       x = "Sample", y = "Rate (mA412/min)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.text.x = element_text(angle = 30, hjust = 1))

ggsave(file.path(output_dir, "technical_replicate_cv.png"), cv_plot,
       width = 9, height = 6, dpi = 300)

cat("--- cv_plot: ggplot object structure ---\n\n")
summary(cv_plot)

cv_plot

--- cv_plot: ggplot object structure ---

data: well, well_type, sample_id, std_nmol, t_start, t_end,
  slope_mOD_min, r2, max_abs_slope_mOD_min, abs_window_is_negative,
  net_change, frac_rising, od_max, flag_decreasing, flag_over_range,
  flag_glitch, glitch_at_min, flag_bg_active, n_flags,
  glitch_in_window, rate_usable, cv_all [24x22]
mapping:  x = ~sample_id, y = ~slope_mOD_min
scales:   colour, shape 
faceting:  <empty> 
-----------------------------------
mapping: colour = ~rate_usable, shape = ~rate_usable 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_jitter 

geom_crossbar: na.rm = FALSE, orientation = NA, width = 0.5
stat_summary: fun.data = NULL, fun = function (x, ...) 
UseMethod("mean"), fun.max = NULL, fun.min = NULL, fun.args = list(), na.rm = FALSE, orientation = NA
position_identity 

mapping: x = ~sample_id, y = Inf, label = ~sprintf("CV %.0f%%", cv_all) 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge

10 CITRATE SYNTHASE ACTIVITY

10.1 Calculation

Following Abcam §10.3:

Sample CS activity = B / (dT x V) x D nmol/min/µL (= mU/µL)

Rearranged for a kinetic slope, where the standard curve converts A412 into nmol of thiol:

  1. Rate (A412/min) = max-increasing sliding-window slope / 1000
  2. Background-corrected rate = sample rate - mean flat background rate
  3. B/dT (nmol/min) = background-corrected rate / standard curve slope (0.05582 A412 per nmol, from the outlier-excluded fit)
  4. Activity (mU/µL) = (B/dT) / V x D, with V = 2 µL and D = 1
  5. Protein-normalized (mU/mg protein) = activity x homogenate_volume_uL (350 µL, measured) / total protein (mg) — recovers total enzyme units in the homogenate, then divides by the total protein extracted from that homogenate (note 3), rather than by tissue input mass
cs_activity <- calculate_cs_activity(well_rates, protein_by_sample, background_per_sample,
                                      std_slope, std_nmol_max, assay_params, plate_long)

cat("Samples with a computed activity:", nrow(cs_activity), "/", nrow(protein_by_sample), "\n")
cat("Activity range (mU/uL):",
    paste(sprintf("%.3f", range(cs_activity$activity_mU_per_uL)), collapse = " - "), "\n")
cat("Activity range (mU/mg protein):",
    paste(sprintf("%.3f", range(cs_activity$activity_mU_per_mg_protein)), collapse = " - "), "\n")

cat("\n--- cs_activity: per-sample citrate synthase activity ---\n\n")
str(cs_activity)
Samples with a computed activity: 8 / 8 
Activity range (mU/uL): 0.061 - 0.171 
Activity range (mU/mg protein): 22.418 - 240.387 

--- cs_activity: per-sample citrate synthase activity ---

tibble [8 × 21] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ n_reps_used               : int [1:8] 3 2 3 2 2 2 3 3
 $ mean_rate_mOD_min         : num [1:8] 7.73 9.13 8 8.38 19.25 ...
 $ sd_rate                   : num [1:8] 0.382 0.813 0.436 0.177 2.899 ...
 $ cv_rate                   : num [1:8] 4.94 8.91 5.45 2.11 15.06 ...
 $ family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ individual                : chr [1:8] "01" "02" "03" "04" ...
 $ temperature               : chr [1:8] "ambient" "ambient" "ambient" "ambient" ...
 $ conc_ug_mL                : num [1:8] 876 1039 1900 1129 712 ...
 $ total_protein_mg          : num [1:8] 0.307 0.364 0.665 0.395 0.249 ...
 $ bg_rate_flat              : num [1:8] 0.55 0.475 0.45 0.225 0.15 ...
 $ n_bg_flat                 : int [1:8] 2 2 2 2 2 1 3 3
 $ bg_rate_mOD_min           : num [1:8] 0.55 0.475 0.45 0.225 0.15 ...
 $ corrected_mOD_min         : num [1:8] 7.18 8.65 7.55 8.15 19.1 ...
 $ rate_OD_min               : num [1:8] 0.00718 0.00865 0.00755 0.00815 0.0191 ...
 $ nmol_per_min              : num [1:8] 0.129 0.155 0.135 0.146 0.342 ...
 $ activity_mU_per_uL        : num [1:8] 0.0643 0.0775 0.0676 0.073 0.1711 ...
 $ total_mU_in_homogenate    : num [1:8] 22.5 27.1 23.7 25.6 59.9 ...
 $ activity_mU_per_mg_protein: num [1:8] 73.4 74.6 35.6 64.6 240.4 ...
 $ nmol_in_window            : num [1:8] 5.15 6.2 5.41 5.84 13.69 ...
 $ within_std_range          : logi [1:8] TRUE TRUE TRUE TRUE TRUE TRUE ...

10.2 Results table

results_table <- build_results_table(cs_activity, cv_summary, baseline_per_sample, assay_params)

kable(results_table,
      caption = paste0("Citrate synthase activity, family F05 ambient (mU/mg protein uses ",
                       "measured protein concentration x the ", assay_params$homogenate_volume_uL,
                       " uL homogenization volume)"))

cat("\nInterpretation summary:\n")
print(results_table %>% count(Interpretation, name = "n_samples") %>%
        as.data.frame(), row.names = FALSE)

write.csv(results_table, file.path(output_dir, "citrate_synthase_activity_results.csv"),
          row.names = FALSE)

cat("\n--- results_table: formatted per-sample results written to CSV ---\n\n")
str(results_table)
Sample Family Individual Temperature Protein conc (ug/mL) Total protein (mg) Reps used CV all reps (%) CV used reps (%) Rate (mA412/min) BG rate (mA412/min) Corrected rate (mA412/min) Activity (mU/uL) Activity (mU/mg protein) CV flag Elevated baseline reps Interpretation
F05_01_ambient F05 01 ambient 876.1 0.307 3/3 4.9 4.9 7.73 0.550 7.18 0.0643 73.442 pass 0/3 usable
F05_02_ambient F05 02 ambient 1038.9 0.364 2/3 30.3 8.9 9.13 0.475 8.65 0.0775 74.579 FAIL >15% 0/3 usable
F05_03_ambient F05 03 ambient 1900.4 0.665 3/3 5.4 5.4 8.00 0.450 7.55 0.0676 35.586 pass 0/3 usable
F05_04_ambient F05 04 ambient 1129.4 0.395 2/3 56.5 2.1 8.38 0.225 8.15 0.0730 64.638 FAIL >15% 1/3 caution - some reps compromised
F05_05_ambient F05 05 ambient 711.7 0.249 2/3 68.8 15.1 19.25 0.150 19.10 0.1711 240.387 FAIL >15% 3/3 DO NOT USE - no clean replicate
F05_06_ambient F05 06 ambient 2232.4 0.781 2/3 36.8 24.7 7.45 0.500 6.95 0.0623 27.886 FAIL >15% 1/3 caution - some reps compromised
F05_07_ambient F05 07 ambient 1687.8 0.591 3/3 6.5 6.5 8.92 0.300 8.62 0.0772 45.729 pass 0/3 usable
F05_08_ambient F05 08 ambient 2703.7 0.946 3/3 7.9 7.9 7.02 0.250 6.77 0.0606 22.418 pass 0/3 usable

Citrate synthase activity, family F05 ambient (mU/mg protein uses measured protein concentration x the 350 uL homogenization volume)

Interpretation summary:
                  Interpretation n_samples
 DO NOT USE - no clean replicate         1
 caution - some reps compromised         2
                          usable         5

--- results_table: formatted per-sample results written to CSV ---

tibble [8 × 17] (S3: tbl_df/tbl/data.frame)
 $ Sample                    : chr [1:8] "F05_01_ambient" "F05_02_ambient" "F05_03_ambient" "F05_04_ambient" ...
 $ Family                    : chr [1:8] "F05" "F05" "F05" "F05" ...
 $ Individual                : chr [1:8] "01" "02" "03" "04" ...
 $ Temperature               : chr [1:8] "ambient" "ambient" "ambient" "ambient" ...
 $ Protein conc (ug/mL)      : num [1:8] 876 1039 1900 1129 712 ...
 $ Total protein (mg)        : num [1:8] 0.307 0.364 0.665 0.395 0.249 0.781 0.591 0.946
 $ Reps used                 : chr [1:8] "3/3" "2/3" "3/3" "2/3" ...
 $ CV all reps (%)           : num [1:8] 4.9 30.3 5.4 56.5 68.8 36.8 6.5 7.9
 $ CV used reps (%)          : num [1:8] 4.9 8.9 5.4 2.1 15.1 24.7 6.5 7.9
 $ Rate (mA412/min)          : num [1:8] 7.73 9.13 8 8.38 19.25 ...
 $ BG rate (mA412/min)       : num [1:8] 0.55 0.475 0.45 0.225 0.15 0.5 0.3 0.25
 $ Corrected rate (mA412/min): num [1:8] 7.18 8.65 7.55 8.15 19.1 6.95 8.62 6.77
 $ Activity (mU/uL)          : num [1:8] 0.0643 0.0775 0.0676 0.073 0.1711 ...
 $ Activity (mU/mg protein)  : num [1:8] 73.4 74.6 35.6 64.6 240.4 ...
 $ CV flag                   : chr [1:8] "pass" "FAIL >15%" "pass" "FAIL >15%" ...
 $ Elevated baseline reps    : chr [1:8] "0/3" "0/3" "0/3" "1/3" ...
 $ Interpretation            : chr [1:8] "usable" "usable" "usable" "caution - some reps compromised" ...

10.3 Plot activity

# Three-level quality status: a sample whose every replicate has an elevated
# baseline cannot be rescued by replicate exclusion, so it outranks CV as a
# disqualifier and is drawn separately (see note 6).
qual_levels <- c("clean", paste0("CV > ", assay_params$cv_threshold_pct, "%"),
                 "all reps baseline-compromised")
qual_cols   <- setNames(c("#08519c", "#cb181d", "grey60"), qual_levels)

activity_plot_data <- cs_activity %>%
  left_join(cv_summary %>% select(sample_id, cv_all), by = "sample_id") %>%
  left_join(baseline_per_sample %>% select(sample_id, n_elevated, n), by = "sample_id") %>%
  mutate(cv_flag = factor(case_when(
                            n_elevated == n                              ~ qual_levels[3],
                            cv_all > assay_params$cv_threshold_pct       ~ qual_levels[2],
                            TRUE                                        ~ qual_levels[1]),
                          levels = qual_levels),
         label_n = paste0(n_reps_used, "/3 reps",
                          ifelse(n_elevated == n, " - DISQUALIFIED", "")))

activity_plot <- ggplot(activity_plot_data,
                        aes(x = reorder(sample_id, activity_mU_per_uL),
                            y = activity_mU_per_uL, fill = cv_flag)) +
  geom_col(width = 0.65) +
  geom_text(aes(label = label_n), hjust = -0.15, size = 3, colour = "grey25") +
  scale_fill_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  coord_flip(clip = "off") +
  expand_limits(y = max(activity_plot_data$activity_mU_per_uL) * 1.25) +
  labs(title = "Citrate synthase activity, M. gigas ctenidia, family F05 ambient",
       subtitle = paste0("Background-corrected, mean of usable technical replicates.\n",
                         "Red = replicate CV above ", assay_params$cv_threshold_pct,
                         "%. Grey = every replicate had an elevated starting A412,\n",
                         "so the apparent high activity is a baseline artifact, not enzyme."),
       x = NULL, y = "CS activity (mU/uL homogenate)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

ggsave(file.path(output_dir, "citrate_synthase_activity.png"), activity_plot,
       width = 10, height = 6, dpi = 300)

cat("--- activity_plot: ggplot object structure ---\n\n")
summary(activity_plot)

activity_plot

--- activity_plot: ggplot object structure ---

data: sample_id, n_reps_used, mean_rate_mOD_min, sd_rate, cv_rate,
  family, individual, temperature, conc_ug_mL, total_protein_mg,
  bg_rate_flat, n_bg_flat, bg_rate_mOD_min, corrected_mOD_min,
  rate_OD_min, nmol_per_min, activity_mU_per_uL,
  total_mU_in_homogenate, activity_mU_per_mg_protein, nmol_in_window,
  within_std_range, cv_all, n_elevated, n, cv_flag, label_n [8x26]
mapping:  x = ~reorder(sample_id, activity_mU_per_uL), y = ~activity_mU_per_uL, fill = ~cv_flag
scales:   fill 
faceting:  <empty> 
-----------------------------------
geom_col: na.rm = FALSE, just = 0.5, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_stack 

mapping: label = ~label_n 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge 

mapping: y = ~y 
geom_blank: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

10.4 Protein-normalized activity

# Clean samples only: the trend test must not be driven by wells whose rate is
# a baseline artifact, so the fit is restricted to samples clean on every check.
norm_clean <- activity_plot_data %>% filter(cv_flag == qual_levels[1])

norm_plot <- ggplot(activity_plot_data,
                    aes(x = total_protein_mg, y = activity_mU_per_mg_protein)) +
  geom_smooth(data = norm_clean, method = "lm", se = TRUE,
              colour = "grey35", fill = "grey85", linewidth = 0.6) +
  geom_point(aes(colour = cv_flag), size = 3.5) +
  geom_text(aes(label = individual), vjust = -1, size = 3, colour = "grey25") +
  scale_colour_manual(values = qual_cols, drop = FALSE, name = "Well quality") +
  labs(title = "Protein-normalized CS activity vs. total extracted protein",
       subtitle = paste0("Point labels = individual ID. Total protein = measured protein ",
                         "concentration x the same\n", assay_params$homogenate_volume_uL,
                         " uL homogenization volume used for every sample. Grey line = fit to ",
                         "clean\nsamples only; a negative slope would indicate the normalization ",
                         "is not protein-independent."),
       x = "Total extracted protein (mg)", y = "CS activity (mU/mg protein)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "bottom")

protein_trend <- cor.test(activity_plot_data$total_protein_mg, activity_plot_data$activity_mU_per_mg_protein)
protein_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_mg_protein)
uL_trend_clean <- cor.test(norm_clean$total_protein_mg, norm_clean$activity_mU_per_uL)

ggsave(file.path(output_dir, "citrate_synthase_activity_normalized.png"), norm_plot,
       width = 10, height = 6, dpi = 300)

cat("Correlation of mU/mg protein with total extracted protein\n")
cat(sprintf("  all %d samples      : r = %+.3f, p = %.4f\n",
            nrow(activity_plot_data), protein_trend$estimate, protein_trend$p.value))
cat(sprintf("  clean samples only  : r = %+.3f, p = %.4f  (n = %d)\n",
            protein_trend_clean$estimate, protein_trend_clean$p.value, nrow(norm_clean)))
cat(sprintf("  clean-sample mU/mg protein range: %.3f to %.3f (%.1f-fold)\n",
            min(norm_clean$activity_mU_per_mg_protein), max(norm_clean$activity_mU_per_mg_protein),
            max(norm_clean$activity_mU_per_mg_protein) / min(norm_clean$activity_mU_per_mg_protein)))
cat(sprintf("  clean-sample mU/uL range: %.4f to %.4f (%.1f-fold)\n",
            min(norm_clean$activity_mU_per_uL), max(norm_clean$activity_mU_per_uL),
            max(norm_clean$activity_mU_per_uL) / min(norm_clean$activity_mU_per_uL)))
cat(sprintf("\nArtifact check: r(total protein, mU/uL raw rate) = %+.3f, p = %.4f  (n = %d)\n",
            uL_trend_clean$estimate, uL_trend_clean$p.value, nrow(norm_clean)))
cat("If mU/uL itself trends with total protein while mU/mg protein trends oppositely,\n",
    "the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate\n",
    "rather than a biological difference in specific activity.\n", sep = "")

cat("\n--- protein_trend_clean: correlation test on clean samples ---\n\n")
str(protein_trend_clean, max.level = 1, give.attr = FALSE)

cat("\n--- norm_clean: clean samples used for the trend fit ---\n\n")
str(norm_clean)

norm_plot

Correlation of mU/mg protein with total extracted protein
  all 8 samples      : r = -0.695, p = 0.0555
  clean samples only  : r = -0.978, p = 0.0215  (n = 4)
  clean-sample mU/mg protein range: 22.418 to 73.442 (3.3-fold)
  clean-sample mU/uL range: 0.0606 to 0.0772 (1.3-fold)

Artifact check: r(total protein, mU/uL raw rate) = -0.274, p = 0.7259  (n = 4)
If mU/uL itself trends with total protein while mU/mg protein trends oppositely,
the mU/mg protein trend is partly/wholly the arithmetic of dividing by a covariate
rather than a biological difference in specific activity.

--- protein_trend_clean: correlation test on clean samples ---

List of 9
 $ statistic  : Named num -6.7
 $ parameter  : Named int 2
 $ p.value    : num 0.0215
 $ estimate   : Named num -0.978
 $ null.value : Named num 0
 $ alternative: chr "two.sided"
 $ method     : chr "Pearson's product-moment correlation"
 $ data.name  : chr "norm_clean$total_protein_mg and norm_clean$activity_mU_per_mg_protein"
 $ conf.int   : num [1:2] -1 -0.291

--- norm_clean: clean samples used for the trend fit ---

tibble [4 × 26] (S3: tbl_df/tbl/data.frame)
 $ sample_id                 : chr [1:4] "F05_01_ambient" "F05_03_ambient" "F05_07_ambient" "F05_08_ambient"
 $ n_reps_used               : int [1:4] 3 3 3 3
 $ mean_rate_mOD_min         : num [1:4] 7.73 8 8.92 7.02
 $ sd_rate                   : num [1:4] 0.382 0.436 0.584 0.558
 $ cv_rate                   : num [1:4] 4.94 5.45 6.55 7.95
 $ family                    : chr [1:4] "F05" "F05" "F05" "F05"
 $ individual                : chr [1:4] "01" "03" "07" "08"
 $ temperature               : chr [1:4] "ambient" "ambient" "ambient" "ambient"
 $ conc_ug_mL                : num [1:4] 876 1900 1688 2704
 $ total_protein_mg          : num [1:4] 0.307 0.665 0.591 0.946
 $ bg_rate_flat              : num [1:4] 0.55 0.45 0.3 0.25
 $ n_bg_flat                 : int [1:4] 2 2 3 3
 $ bg_rate_mOD_min           : num [1:4] 0.55 0.45 0.3 0.25
 $ corrected_mOD_min         : num [1:4] 7.18 7.55 8.62 6.77
 $ rate_OD_min               : num [1:4] 0.00718 0.00755 0.00862 0.00677
 $ nmol_per_min              : num [1:4] 0.129 0.135 0.154 0.121
 $ activity_mU_per_uL        : num [1:4] 0.0643 0.0676 0.0772 0.0606
 $ total_mU_in_homogenate    : num [1:4] 22.5 23.7 27 21.2
 $ activity_mU_per_mg_protein: num [1:4] 73.4 35.6 45.7 22.4
 $ nmol_in_window            : num [1:4] 5.15 5.41 6.17 4.85
 $ within_std_range          : logi [1:4] TRUE TRUE TRUE TRUE
 $ cv_all                    : num [1:4] 4.94 5.45 6.55 7.95
 $ n_elevated                : int [1:4] 0 0 0 0
 $ n                         : int [1:4] 3 3 3 3
 $ cv_flag                   : Factor w/ 3 levels "clean","CV > 15%",..: 1 1 1 1
 $ label_n                   : chr [1:4] "3/3 reps" "3/3 reps" "3/3 reps" "3/3 reps"

The clean-sample correlation of mU/mg protein with total extracted protein is r = -0.978 (p = 0.0215, n = 4) — as strongly negative as the tissue-weight version of this check was. The difference is in the artifact check: raw mU/µL correlates with total protein at r = -0.274 (p = 0.726), a real (if not clean) inverse trend rather than the flat line seen against tissue weight. That means the mU/mg protein trend is not purely a division artifact the way mU/mg tissue was — part of it reflects that samples with more extracted protein measured a lower raw rate, which a fixed-volume, decelerating-kinetics assay (notes 3–4) cannot distinguish from a genuine difference in specific activity without a dilution series. Both possibilities remain open on this plate; the number should be treated as provisional pending that follow-up, but it is not dismissible as pure arithmetic the way the tissue-weight version was.

11 QC SUMMARY

qc_summary <- build_qc_summary_table(layout_wells, recon$summary, std_curve,
                                      well_diagnostics, well_rates, cv_summary,
                                      baseline_per_sample, pos_control, assay_params)

kable(qc_summary, col.names = c("QC check", "Result"),
      caption = "Quality control summary for this plate")

write.csv(qc_summary, file.path(output_dir, "qc_summary.csv"), row.names = FALSE)

cat("\n--- qc_summary: consolidated QC checks for the run ---\n\n")
str(qc_summary)
QC check Result
Occupied wells in layout 69
Layout wells missing from absorbance CSV 0
Max disagreement between absorbance CSV and full report 0
Standard concentrations with replicate CV > threshold 0, 8, 16, 24
Standard wells flagged as outliers 8/18 (E6, E2, E12, E8, E10, F3, E7, F5)
Standard curve R^2 (outlier-excluded, replicate level) 0.9678
Standard curve R^2 (all wells, replicate level) 0.8042
Standard wells that LOST signal over the run 16/18
Standards above photometric linearity ceiling 32, 40
Background control wells behaving as active reactions 7/24
Sample wells with a decreasing trace 3/24
Sample wells with an elevated starting A412 5/24 (A12, B1, B2, B3, B5)
Samples with ALL THREE replicates baseline-compromised 1/8 (F05_05_ambient)
Sample wells with a read glitch 5/24
Sample wells usable for rate extraction 20/24
Samples with technical CV > threshold (all reps) 4/8
Samples with technical CV > threshold (usable reps) 2/8
Positive control replicates rising and linear 3/3

Quality control summary for this plate

--- qc_summary: consolidated QC checks for the run ---

'data.frame':   18 obs. of  2 variables:
 $ check: chr  "Occupied wells in layout" "Layout wells missing from absorbance CSV" "Max disagreement between absorbance CSV and full report" "Standard concentrations with replicate CV > threshold" ...
 $ value: chr  "69" "0" "0" "0, 8, 16, 24" ...

11.1 Auto-generated findings

The bullets below are generated programmatically from the QC/CV/anomaly objects above by generate_qc_findings(), so the same logic that flags a problem here is guaranteed to be the logic that flagged it upstream.

qc_findings <- generate_qc_findings(well_diagnostics, well_rates, cv_summary,
                                     baseline_per_sample, std_curve, recon$summary,
                                     assay_params)

cat(paste0("- ", qc_findings, "\n"), sep = "\n")
- **Kinetic sources agree.** absorbance CSV and full report agree exactly on all 1449 shared well x timepoint readings; all layout wells are present.

- **Decreasing kinetics.** 3 sample well(s) end lower than (or equal to) their starting A412 and are excluded from rate extraction: A12, B3, B5.

- **Read glitches.** 5 sample well(s) contain a single read-to-read step inconsistent with its neighbours: A12, A6, B1, B2, B5 (only disqualifying if it falls inside the fitted rate window).

- **Elevated starting absorbance.** 5 sample well(s) start above the baseline threshold (A412 > 0.35), indicating pre-existing thiol/contamination before the reaction began: A12, B1, B2, B3, B5.

- **Sample(s) with no clean replicate.** Every replicate of F05_05_ambient has an elevated starting A412; these samples have no usable measurement on this plate regardless of CV.

- **Over-range wells.** 7 well(s) exceed the photometric linearity ceiling (A412 > 1.5): E10, F1, F2, F3, F4, F5, F6.

- **Background control wells behaving as active reactions.** 7 of 24 background wells show a starting A412 or drift inconsistent with a flat, inactive well: C12, C2, C5, C7, D1, D4, D5.

- **GSH standard curve replicate imprecision.** 4 of 6 concentrations exceed 15% CV: 0, 8, 16, 24 nmol/well.

- **Standard outlier wells.** 8 of 18 standard wells deviate from their triplicate median by more than 0.15 A412: E6, E2, E12, E8, E10, F3, E7, F5.

- **Standards above the photometric linear range.** 32, 40 nmol/well read above A412 1.5; the calibration is trustworthy only below these concentrations.

- **Standard signal decay.** 16 of 18 standard wells lost signal over the run (TNB instability); the standard curve is read at t = 0.

- **Standard curve fit (outlier-excluded):** slope = 0.05582 A412/nmol, R^2 = 0.9678 (all-wells R^2 = 0.8042).

- **Technical replicate CV > 15% (all replicates).** 4 of 8 samples: F05_02_ambient (30%), F05_04_ambient (56%), F05_05_ambient (69%), F05_06_ambient (37%).

- **Technical replicate CV > 15% persists after excluding flagged replicates.** F05_05_ambient (15%, n=2), F05_06_ambient (25%, n=2).

12 SUMMARY

12.1 Anomalies found

  1. absorbance-*.csv originally omitted plate columns 10–12 — now resolved. An earlier export of this file covered only 54 of the 69 occupied wells (missing F05_04, F05_08, their background controls, and the whole 24 nmol GSH standard). The file has been re-exported from Gen5 and now covers all 69 wells; the reconcile-sources chunk confirms every layout well is present and that the re-export agrees exactly with full_report-*.txt on all 1,449 shared readings. No wells were recovered from the full report in this version of the analysis.

  2. Background control wells are not behaving as background controls. Of 24 background wells, roughly a third start at high A412 (0.24–0.82 vs. ~0.05 for well-behaved wells) and then decrease steeply — up to -19 mA412/min. Background Control Mix contains no CS Substrate Mix, so these wells cannot support a CS reaction and must read flat. Six of the eight background triplicates contain at least one such well, and they appear in a near-regular pattern across the plate. One well (C2) instead increases at 11.6 mA412/min, i.e. it behaves like a full reaction well. The most consistent explanation is that Reaction Mix was dispensed into some background wells instead of Background Control Mix. Because of this, background correction here uses only the flat replicates, and the flat-well background rate is ~0.3 mA412/min — about 3% of the median sample signal, so background correction is a minor term regardless.

  3. The GSH standard curve failed QC. Replicate CVs run from 6% to 126%, and five of six concentrations exceed the 15% threshold. The replicate-level R2 is only 0.80. Two wells are grossly discrepant from their triplicate medians (E6 at the 8 nmol point, +0.80 A412; E2 at the 0 nmol point, +0.77 A412) — E2 reads 0.871 while its siblings E1 and E3 read 0.100 and 0.095. This is the dominant source of uncertainty in every reported activity value, since the standard curve slope scales all of them.

  4. The top two standards are above the photometric linear range. The 32 and 40 nmol standards read A412 1.7–2.2. Response per nmol is essentially flat between them (net A412 1.61 vs. 1.72 for a 25% concentration increase), i.e. the curve is saturating. The calibration is trustworthy only over roughly 0–24 nmol.

  5. Every standard well lost signal over the run (18/18 wells, drift -0.001 to -0.74 A412). GSH + DTNB is a stoichiometric endpoint, so standards should be stable; this indicates TNB2- instability under the run conditions. The standard curve is therefore read at t = 0, before appreciable decay.

  6. Decreasing kinetics in sample wells. Wells A12, B3 and B5 end lower than they started, and A6 rises by only 0.041 A412 total against ~0.16 for its siblings. These wells are excluded from rate extraction. Note that Gen5’s own Max V [412] column reports large negative rates for these wells (e.g. -19.7 for A12, -11.5 for B3) because Gen5 selects the window of maximum absolute slope; those numbers are not activities and must not be used as such.

  7. Read glitches. Eleven wells contain a single read-to-read step that disagrees with the steps on either side of it by more than 0.02 A412 — among sample wells: A6 (+0.040 at 30 min, after the trace had gone flat), A12 (-0.036 at 2 min), B1 (-0.035 at 4 min), B2 (-0.041 at 4 min) and B5 (+0.031 at 2 min); also D4, E6, E7, E10, E11 and F1. These are instrument artifacts rather than kinetics. A glitch only disqualifies a well’s rate when it falls inside the fitted window, which is why B1 and B2 retain usable rates (their glitch at 4 min precedes their 10–18 min window) while A6 and B5 do not.

  8. Elevated starting absorbance is the pattern underlying most of the above. Five sample wells (A12, B1, B2, B3, B5) and seven background wells start far above the clean-well baseline of A412 0.13–0.16, and these are precisely the wells that decline or misbehave. TNB present at t = 0 means the reaction had already run, or the well was contaminated, before the first read. The affected wells cluster in the right-hand columns of rows A–B and scattered positions in C–D, which points to a dispensing or timing problem rather than random error. Most consequentially, all three F05_05 replicates are affected, which disqualifies that sample entirely.

  9. Layout label problem. F05_05_ambient_BG (D1–D3) is spelled citrate_synthasse. Handled in code, but the layout file should be corrected.

12.2 Technical replicate precision

Across all three replicates, four of eight samples exceed the 15% CV threshold: F05_05 (69%), F05_04 (56%), F05_06 (37%) and F05_02 (30%). The other four — F05_01, F05_03, F05_07, F05_08 — sit at 4.9%, 5.4%, 6.5% and 7.9% respectively.

In every failing case the CV is driven by a single bad well, not by uniform imprecision. Excluding that one flagged replicate drops F05_02 to 8.9% (excluding A6), F05_04 to 2.1% (A12) and F05_05 to 15.1% (B3). F05_06 is the exception: excluding B5 still leaves 24.7%, because its remaining two replicates (B4 at 6.15 and B6 at 8.75 mA412/min) genuinely disagree. Every sample retains at least two usable replicates.

So two samples remain above threshold after QC exclusion: F05_06 (24.7%, n = 2) and F05_05 (15.1%, n = 2, marginal).

F05_05 should not be used at all, and its CV is not the reason. It reports 19.3 mA412/min — 2.2x the next-highest sample — but all three of its replicates start at elevated A412 (0.65, 0.67, 0.93 versus 0.13–0.16 for clean wells). Its two “usable” wells (B1, B2) are the two steepest traces on the plate and both rise from an already-contaminated baseline before turning over, and B3 declines outright. A well that begins with TNB already present is not measuring citrate synthase from zero, so the apparent high activity is an artifact of the starting condition, not enzymology. Excluding one replicate on CV grounds cannot repair a sample in which no replicate is clean.

12.3 Assay validity

The positive control worked: all three replicates rise linearly (R2 > 0.999) at ~2.8 mA412/min with 2% CV. The core reaction chemistry and the reader were functioning. The problems on this plate are in the standards and background wells, not in the CS reaction itself.

Sample rates in clean wells are also well-behaved and mutually consistent (~7–9 mA412/min for most individuals, R2 > 0.99 over the first 8 min), which is why per-sample activity estimates are reported despite the standard curve problems.

12.4 Recommendation

Treat the activity values here as provisional. The rate measurements are sound for the wells that passed QC, but they are calibrated through a standard curve that failed its own QC, so absolute activities carry an uncertainty that the reported precision does not capture. Before these numbers are used downstream:

  1. Run a dilution series on one homogenate — the highest priority item. Assay a single sample neat, 1:2, 1:5 and 1:10 and plot rate against dilution. Measured mU/µL correlates with total extracted protein at r = -0.27 (p = 0.73) across a 3-fold range of protein input on this plate — a real but noisy inverse trend, not the flat line seen against tissue weight. A dilution series is the only way to tell whether that reflects the assay running outside its linear range (aliquot volume, substrate depletion) or a genuine difference in specific activity between individuals; nothing on this list should be treated as settled until it is run.
  2. Re-run the GSH standard curve, keeping the top standard at or below 24 nmol/well so all points stay within photometric linearity, and read it immediately.
  3. Resolve the background control mix-up — confirm which wells received Background Control Mix vs. Reaction Mix before the next plate.
  4. Re-assay F05_05 — mandatory, not optional. All three replicates are baseline-compromised, so this sample has no usable measurement on this plate regardless of what the CV column says. Do not carry its 240 mU/mg protein into any downstream comparison. Re-assay F05_06 as well (24.7% CV on two genuinely disagreeing replicates).
  5. Check the dispensing/plating step for the elevated-baseline wells (A12, B1, B2, B3, B5 — all in the bottom-right region of rows A–B, plus background wells C2, C5, C7, C12, D1, D4, D5). The spatial clustering suggests a systematic cause — carryover on a multichannel tip, DTNB added before the plate reached the reader, or a delay between reagent addition and the first read — rather than random pipetting error.
  6. Protein normalization controls for extraction differences, but not for whatever is driving the mU/uL-vs-protein trend. Normalizing to total extracted protein (rather than tissue weight) already removes sample-to-sample variation in how much soluble material a given tissue input yielded. It does not, by itself, explain why raw mU/µL trends downward with more extracted protein (recommendation 1) — that requires the dilution series, not a change in normalization basis.
  7. Re-export the absorbance CSV covering all 12 plate columns. Done — absorbance-*.csv now covers the full plate; see anomaly

12.5 What can be used from this plate

Four samples are clean on every check — F05_01, F05_03, F05_07 and F05_08, all with no baseline-compromised replicates and CV < 8%. Their relative ranking is more robust than their absolute values, since all four share the same calibration.

The instructive result is how little they differ: 6.8, 7.6, 8.6 and 6.8 mA412/min background-corrected — a spread of roughly 25%, against a technical CV of 5–8% within samples. Once the plate’s one apparent outlier (F05_05) is recognized as a baseline artifact rather than a high-activity individual, no striking within-family variation remains at ambient temperature. Two implications for the wider design:

  • The mU/mg protein values should be treated as provisional, not discarded outright. The standard curve QC failure (three of six concentrations with CV > 15%, 8/18 wells flagged) is one source of uncertainty affecting every reported value equally. Separately, mU/mg protein correlates strongly negatively with total extracted protein (r = -0.98, clean samples), but — unlike the tissue-weight version of this plate — the underlying raw rate (mU/µL) also correlates with extracted protein (r = -0.27), so the mU/mg protein trend is not pure division artifact. Whether it reflects real specific-activity differences between individuals or an assay linear-range effect is unresolved without the dilution series.
  • The assay needs a linear-range check before more plates are run. The dilution series is cheap — one homogenate, four dilutions — and it determines whether the deceleration seen in every trace, and its apparent relationship to protein content, is substrate depletion, aliquot overload, or extraction saturation rather than biology. Running more families before that is answered risks generating plates with an uninterpretable normalization.
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase ABCAM https://robertslab.github.io/sams-notebook/posts/2026/2026-08-11-Citrate-Synthase-Assay---SORMI-June-2026-M.gigas-Family-5-Ambient-Ctenidia/ Tue, 11 Aug 2026 07:00:00 GMT Homogenization - SORMI June 2026 M.gigas Ctenidia from Families 5 and 7 for Citrate Synthase Assay Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-11-Homogenization---SORMI-June-2026-M.gigas-Ctenidia-from-Families-5-and-7-for-Citrate-Synthase-Assay/

INTRO

Families 5 & 7 from the June 2026 SORMI experiment (GitHub repo) were selected for citrate synthase assay (selection by Steven). Ctenidia samples from these families were homogenized in preparation for the Citrate Synthase Assay Kit (ABCAM; cat: ab239712).

MATERIALS & METHODS

Eight ctenidia samples (1-8) from each treatment (ambient and 36°C) in Families 5 and 7 were selected. Samples were stored on dried ice until homogenization. Pieces of ctenidia were broken off and weighed to the nearest 0.1 mg. Tissue was transferred to 1.5mL Safe-Lock tubes (Eppendorf) containing 350 µL of Assay Buffer 7 and ~100µL of 0.5mm glass beads (Research Products International). Samples were homogenized in a pre-/cooled (dry ice) Bullet Blender 5E Gold+ using the 1.5mL adapters. Homogenization was performed for 10 minutes at Speed 12. After homogenization, samples were stored at -80°C until further processing.

Note

A homogenization volume of 350 µL (instead of the manufacturer’s recommended 100 µL) was used based off of prior work in our lab with this kit, performed by Matt George (link to the paper - requires institutional access).

Below is a list of the samples and the weights of ctenidia used for homogenization:

Sample Weight(mg)
F05_01_ambient 10.1
F05_02_ambient 17.5
F05_03_ambient 22.3
F05_04_ambient 14
F05_05_ambient 11.5
F05_06_ambient 30.2
F05_07_ambient 25.8
F05_08_ambient 30.6
F05_01_36C 12.8
F05_02_36C 6.2
F05_03_36C 10.3
F05_04_36C 12.8
F05_05_36C 22.7
F05_06_36C 18.4
F05_07_36C 11.3
F05_08_36C 13.8
F07_01_ambient 7.3
F07_02_ambient 19.4
F07_03_ambient 19.7
F07_04_ambient 10.7
F07_05_ambient 8.3
F07_06_ambient 12.8
F07_07_ambient 20.6
F07_08_ambient 10.9
F07_01_36C 9
F07_02_36C 9.5
F07_03_36C 18.3
F07_04_36C 16.9
F07_05_36C 10.5
F07_06_36C 9.7
F07_07_36C 12.5
F07_08_36C 9.1
]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas citrate synthase homogenization https://robertslab.github.io/sams-notebook/posts/2026/2026-08-11-Homogenization---SORMI-June-2026-M.gigas-Ctenidia-from-Families-5-and-7-for-Citrate-Synthase-Assay/ Tue, 11 Aug 2026 07:00:00 GMT Trimming - Andy Dittman RNA-seq Data Using fastp FastQC MultiQC on Hyak Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-10-Trimming---Andy-Dittman-RNA-seq-Data-Using-fastp-FastQC-MultiQC-on-Hyak/

INTRO

After running an initial FastQC on the raw reads on 20260806 (notebook entry), I thought it would be prudent to run the reads through some adapter and quality trimming. I ran fastp to trim the reads (removed adapters and low-quality bases, 15bp from 5’ end of each read, removed poly-G and poly-A tails), and then ran FastQC and MultiQC on the trimmed reads to assess the quality of the trimmed data. See CODE section below for specifics.

The trimming and quality assessment was run on Hyak using our Apptainer image: srlab-R4.4-bioinformatics-container-c3d3116.sif.

RESULTS

Output directory:

MultiQC report (HTML):

DISCUSSION

Trimming with fastp went quickly and smoothly, and the resulting trimmed FastQs were of good quality, as assessed by FastQC and MultiQC. After trimming, the per base sequence content is much more consistent across the read length.

Will let Andy know.

Code is below.

1 SETUP

1.1 Libraries

library(knitr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2 VARIABLES

# DIRECTORIES
raw_reads_dir <- "/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1/"
output_dir <- "/mmfs1/gscratch/scrubbed/samwhite/outputs/20260806-Fastp-FastQC-MultiQC-dittman-grc-rnaseq"

# FILES
fastq_pattern="*.fastq.gz"
R1_fastq_pattern="*_R1_*.fastq.gz"
R2_fastq_pattern="*_R2_*.fastq.gz"
trimmed_fastq_pattern="*fastp-trim.fq.gz"



# SETTINGS
threads <- "50"


# Export these as environment variables for bash chunks.
Sys.setenv(
  fastq_pattern = fastq_pattern,
  raw_reads_dir = raw_reads_dir,
  R1_fastq_pattern = R1_fastq_pattern,
  R2_fastq_pattern = R2_fastq_pattern,
  trimmed_fastq_pattern = trimmed_fastq_pattern,
  output_dir = output_dir,
  threads = threads
)

3 Fastp Trimming

fastp [@chen2023] is set to auto-detect Illumina adapters, as well as trim the first 15bp from each read, as past experience shows these first 15bp are more inconsistent than the remainder of the read length.

Expects filenames like: lib_212412_sample_1703783_S120_L002_R2_001.fastq.gz

# Make output directories, if it doesn't exist
mkdir --parents "${output_dir}"

# Change to raw reads directory
cd "${raw_reads_dir}"

# Create arrays of fastq R1 files and sample names
for fastq in ${R1_fastq_pattern}
do
  fastq_array_R1+=("${fastq}")
  R1_names_array+=("$(echo "${fastq}" | awk -F"_" -v OFS="_" '{print $1, $2, $3, $4, $5, $6}')")
done

# Create array of fastq R2 files
for fastq in ${R2_fastq_pattern}
do
  fastq_array_R2+=("${fastq}")
  R2_names_array+=("$(echo "${fastq}" | awk -F"_" -v OFS="_" '{print $1, $2, $3, $4, $5, $6}')")
done

# Create list of fastq files used in analysis
# Create MD5 checksum for reference
if [ ! -f "${output_dir}"/raw-fastq-checksums.md5 ]; then
for fastq in *.gz
  do
    md5sum ${fastq} >> "${output_dir}"/raw-fastq-checksums.md5
  done
fi

# Run fastp on files
# Adds JSON report output for downstream usage by MultiQC
for index in "${!fastq_array_R1[@]}"
do
  R1_sample_name=$(echo "${R1_names_array[index]}")
  R2_sample_name=$(echo "${R2_names_array[index]}")
  fastp \
  --in1 ${fastq_array_R1[index]} \
  --in2 ${fastq_array_R2[index]} \
  --detect_adapter_for_pe \
  --trim_poly_g \
  --trim_poly_x \
  --trim_front1 15 \
  --trim_front2 15 \
  --thread ${threads} \
  --html "${output_dir}"/"${R1_sample_name}".fastp-trim.report.html \
  --json "${output_dir}"/"${R1_sample_name}".fastp-trim.report.json \
  --out1 "${output_dir}"/"${R1_sample_name}"_R1_001.fastp-trim.fq.gz \
  --out2 "${output_dir}"/"${R2_sample_name}"_R2_001.fastp-trim.fq.gz \
  2>> "${output_dir}"/fastp.stderr

  # Generate md5 checksums for newly trimmed files
  cd "${output_dir}"
  md5sum "${R1_sample_name}"_R1_001.fastp-trim.fq.gz > "${R1_sample_name}"_R1_001.fastp-trim.fq.gz.md5
  md5sum "${R2_sample_name}"_R2_001.fastp-trim.fq.gz > "${R2_sample_name}"_R2_001.fastp-trim.fq.gz.md5
  cd -
done
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1

4 FASTQC/MULTIQC

Uses --cl-config "sp: { fastp: { fn: '*report.json' } }" to update the MultiQC search pattern for the fastp module.

# Make output directory if it doesn't exist
mkdir --parents "${output_dir}"

############ RUN FASTQC ############


# Create array of trimmed FastQs
trimmed_fastqs_array=(${output_dir}/${trimmed_fastq_pattern})

# Pass array contents to new variable as space-delimited list
trimmed_fastqc_list=$(echo "${trimmed_fastqs_array[*]}")

echo "Beginning FastQC on trimmed reads..."
echo ""

# Run FastQC
### NOTE: Do NOT quote trimmed_fastqc_list
fastqc \
--threads ${threads} \
--outdir ${output_dir} \
--quiet \
${trimmed_fastqc_list}

echo "FastQC on trimmed reads complete!"
echo ""

############ END FASTQC ############

############ RUN MULTIQC ############
echo "Beginning MultiQC on trimmed FastQC..."
echo ""

multiqc ${output_dir} \
--cl-config "sp: { fastp: { fn: '*report.json' } }" \
--interactive \
-o ${output_dir}

echo ""
echo "MultiQC on trimmed FastQs complete."
echo ""

############ END MULTIQC ############

echo "Removing FastQC zip files."
echo ""
rm ${output_dir}/*.zip
echo "FastQC zip files removed."
echo ""
Beginning FastQC on trimmed reads...

application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
FastQC on trimmed reads complete!

Beginning MultiQC on trimmed FastQC...


/// MultiQC 🔍 v1.24.1

     version_check | MultiQC Version v1.35 now available!
       file_search | Search path: /mmfs1/gscratch/scrubbed/samwhite/outputs/20260806-Fastp-FastQC-MultiQC-dittman-grc-rnaseq
             fastp | Found 96 reports
            fastqc | Found 192 reports
     write_results | Data        : /mmfs1/gscratch/scrubbed/samwhite/outputs/20260806-Fastp-FastQC-MultiQC-dittman-grc-rnaseq/multiqc_data
     write_results | Report      : /mmfs1/gscratch/scrubbed/samwhite/outputs/20260806-Fastp-FastQC-MultiQC-dittman-grc-rnaseq/multiqc_report.html
           multiqc | MultiQC complete

MultiQC on trimmed FastQs complete.

Removing FastQC zip files.

FastQC zip files removed.
]]> 2026 fastp fastqc multiqc hyak RNA-seq trimming https://robertslab.github.io/sams-notebook/posts/2026/2026-08-10-Trimming---Andy-Dittman-RNA-seq-Data-Using-fastp-FastQC-MultiQC-on-Hyak/ Mon, 10 Aug 2026 07:00:00 GMT FastQC MultiQC - Andy Dittman Remaining RNA-seq Data on Hyak Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-06-FastQC-MultiQC---Andy-Dittman-Remaining-RNA-seq-Data-on-Hyak/

INTRO

The remainder of RNA-seq data for Andy Dittman at NOAA came in on 20260715 (Notebook entry). The data is the remainder of the RNA-seq data, after an initial run of a subset of samples to see how the data looked and to make sure it was of good quality. The initial GitHub Issue indicated that the RNA Integrity Number (RIN) for these samples was not ideal, however, the initial FastQC/MultiQC (notebook entry) and subsequent trimming (notebook entry) showed the data was good. I ran FastQC and MultiQC on these new raw reads to assess their quality.

Analysis was run on Hyak using the following Roberts Lab Apptainer image:

  • srlab-R4.4-bioinformatics-container-c3d3116.sif (built on 20260410)

RESULTS

Output directory:

MultiQC Report (HTML):

Overall, the FastQC and MultiQC reports indicate that the raw reads are of good quality, with quality being consistent for what we usually see with RNA-seq data. Will proceed with quality trimming and adapter removal using FastP, and then re-run FastQC and MultiQC on the trimmed reads.

Code is below.

1 SETUP

1.1 Libraries

library(knitr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2 VARIABLES

# DIRECTORIES
raw_reads_dir <- "/mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1"
output_dir <- "/mmfs1/gscratch/scrubbed/samwhite/outputs/20260806-FastQC-MultiQC-dittman-grc-rnaseq/"

# FILES
fastq_pattern="*.fastq.gz"

# SETTINGS
threads <- "50"


# Export these as environment variables for bash chunks.
Sys.setenv(
  fastq_pattern = fastq_pattern,
  raw_reads_dir = raw_reads_dir,
  output_dir = output_dir,
  threads = threads
)

3 FASTQC/MULTIQC

# Make output directory if it doesn't exist
mkdir --parents "${raw_reads_dir}"

############ RUN FASTQC ############


# Create array of trimmed FastQs
raw_fastqs_array=(${raw_reads_dir}/${fastq_pattern})

# Pass array contents to new variable as space-delimited list
raw_fastqc_list=$(echo "${raw_fastqs_array[*]}")

echo "Beginning FastQC on raw reads..."
echo ""

# Run FastQC
### NOTE: Do NOT quote raw_fastqc_list
fastqc \
--threads ${threads} \
--outdir ${raw_reads_dir} \
--quiet \
${raw_fastqc_list}

echo "FastQC on raw reads complete!"
echo ""

############ END FASTQC ############

############ RUN MULTIQC ############
echo "Beginning MultiQC on raw FastQC..."
echo ""

multiqc ${raw_reads_dir} \
--interactive \
-o ${raw_reads_dir}

echo ""
echo "MultiQC on raw FastQs complete."
echo ""

############ END MULTIQC ############

echo "Removing FastQC zip files."
echo ""
rm ${raw_reads_dir}/*.zip
echo "FastQC zip files removed."
echo ""
Beginning FastQC on raw reads...

application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
application/gzip
FastQC on raw reads complete!

Beginning MultiQC on raw FastQC...


/// MultiQC 🔍 v1.24.1

     version_check | MultiQC Version v1.35 now available!
       file_search | Search path: /mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1
            fastqc | Found 155 reports
     write_results | Data        : /mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1/multiqc_data
     write_results | Report      : /mmfs1/gscratch/scrubbed/samwhite/data/dittman_grc_rnaseq_1/multiqc_report.html
           multiqc | MultiQC complete

MultiQC on raw FastQs complete.

Removing FastQC zip files.

FastQC zip files removed.
]]> 2026 FastQC MultiQC RNA-seq hyak https://robertslab.github.io/sams-notebook/posts/2026/2026-08-06-FastQC-MultiQC---Andy-Dittman-Remaining-RNA-seq-Data-on-Hyak/ Thu, 06 Aug 2026 07:00:00 GMT Glycogen Analysis - SORMI June 2026 M.gigas Family 1 vs Family 9 Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-06-Glycogen-Analysis---SORMI-June-2026-M.gigas-Family-1-vs-Family-9/

INTRO

This notebook analyzed glycogen content in Magallana gigas (Pacific oyster) ctenidia samples from USDA families 1 and 9, comparing ambient vs. 36°C temperature exposure, results from two assay dates (20260730 and 20260803 (Notebook entries)).

See the BACKGROUND section below for details on the re-assay rationale, the standard-curve QC checks, and the outlier/replicate-exclusion analysis for 1_04_36C.

AI Use Level 4: Substantial AI contribution

Note

This section was knitted from https://github.com/RobertsLab/sormi-assay-development/blob/main/Glycogen/code/Gen5-20260804-mgig-fams1_9-glycogen-glo-combined.Rmd (GitHub).

1 BACKGROUND

Combined glycogen analysis of Magallana gigas (Pacific oyster) ctenidia samples from USDA families 1 and 9, comparing ambient vs. 36°C temperature exposure, using the Glycogen-Glo Assay (Promega) (GitHub; PDF).

This document merges results from two assay dates:

  • 2026-07-30 (Gen5-20260730-mgig-fams1_9-glycogen-glo): the full 32-sample experiment (all family × temperature combinations), samples run at 1:25 dilution on one plate, standard curve + negative control run on a separate plate.

  • 2026-08-03 (Gen5-20260803-mgig-fams1_9-glycogen-glo): a targeted re-assay of 5 samples at 1:200 dilution (~8× higher), run alongside a fresh, same-plate standard curve. Four of these five (1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient) had read above the top standard (20 µg/mL) in the original run and required extrapolation; the fifth (1_04_36C) was in range originally but had an elevated technical-replicate CV (30.1%).

Reconciliation rule applied here: for each of the 32 samples, this document selects whichever available measurement falls within its own run’s standard-curve range (interpolated rather than extrapolated), preferring the more recent run as a tiebreaker if more than one candidate qualifies. In practice this means:

  • 1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient: the 2026-08-03 (1:200) value is used, since it is in-range and the 2026-07-30 value was extrapolated.
  • 1_04_36C: the 2026-07-30 (1:25) value is used per instruction – its 2026-08-03 (1:200) re-assay read below the standard curve’s floor (negative back-calculated concentration, unusable), so the original in-range-but-high-CV value is retained here rather than the unusable re-assay value.
  • All other 28 samples: only the 2026-07-30 value exists and is used as-is.

Full detail on the re-assay rationale, the standard-curve QC checks, and the outlier/replicate-exclusion analysis for 1_04_36C (a Grubbs test found no statistical basis for dropping a replicate, and a general “exclude any replicate outside 1 SD of the triplicate mean” rule was tested and rejected as a mathematical inevitability for n=3, not a data-quality signal) are documented in the two source analyses linked above and are not repeated here.

1.1 Sample naming

Per ../data/raw_luminescence/README.md, plate layout entries follow <sample>-<assay_type>-<tissue_weight>-df.<dilution_factor>, with sample tokens formatted as <family>_<individual>_<temperature>, e.g. 1_06_ambient-glyc-11.0-df.25. Standards are labeled STD-glyc-<concentration> and the negative (buffer-only) control NEG-glyc.

2 SETUP

2.1 Libraries

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Set output directory

output_dir <- "../output/Gen5-20260804-mgig-fams1_9-glycogen-glo-combined"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

3 DATA IMPORT

Raw plate-reader exports (comma-separated, no header, one row per plate row A-H and one column per plate column 1-12) are read separately for each plate, for both assay dates.

data_dir <- "../data/raw_luminescence"

# 2026-07-30: plate-01 (32 samples, triplicate wells), plate-02 (standard curve + neg. control)
layout_0730_plate01 <- read.csv(file.path(data_dir, "layout-Gen5-20260730-mgig-fams1_9-plate-01.csv"),
                                header = FALSE, stringsAsFactors = FALSE)
raw_0730_plate01    <- read.csv(file.path(data_dir, "raw_lum-Gen5-20260730-mgig-fams1_9-plate-01.csv"),
                                header = FALSE)
layout_0730_plate02 <- read.csv(file.path(data_dir, "layout-Gen5-20260730-mgig-fams1_9-plate-02.csv"),
                                header = FALSE, stringsAsFactors = FALSE)
raw_0730_plate02    <- read.csv(file.path(data_dir, "raw_lum-Gen5-20260730-mgig-fams1_9-plate-02.csv"),
                                header = FALSE)

# 2026-08-03: single plate (5 re-assayed samples + fresh standard curve + neg. control)
layout_0803_plate01 <- read.csv(file.path(data_dir, "layout-Gen5-20260803-mgig-fams1_9-plate-01.csv"),
                                header = FALSE, stringsAsFactors = FALSE)
raw_0803_plate01    <- read.csv(file.path(data_dir, "raw_lum-Gen5-20260803-mgig-fams1_9-plate-01.csv"),
                                header = FALSE)

cat("2026-07-30 plate-01 dims:", dim(layout_0730_plate01), "\n")
cat("2026-07-30 plate-02 dims:", dim(layout_0730_plate02), "\n")
cat("2026-08-03 plate-01 dims:", dim(layout_0803_plate01), "\n")
2026-07-30 plate-01 dims: 8 12 
2026-07-30 plate-02 dims: 8 12 
2026-08-03 plate-01 dims: 8 12

4 PLATE RESHAPING

Converts each plate’s row/column layout into long format (one row per well), dropping any well without a sample/standard label (empty wells).

plate_to_long <- function(layout, luminescence, plate_label) {
  n_row <- 8
  n_col <- 12
  out <- expand.grid(plate_row_idx = 1:n_row, plate_col = 1:n_col)
  out$plate        <- plate_label
  out$plate_row    <- LETTERS[out$plate_row_idx]
  out$well         <- sprintf("%s%02d", out$plate_row, out$plate_col)
  out$label        <- trimws(as.character(mapply(function(i, j) layout[i, j],
                                                  out$plate_row_idx, out$plate_col)))
  out$luminescence <- as.numeric(mapply(function(i, j) luminescence[i, j],
                                        out$plate_row_idx, out$plate_col))
  out <- out[order(out$plate_row_idx, out$plate_col), ]
  out[out$label != "" & !is.na(out$label), ]
}
plate_0730_01_long <- plate_to_long(layout_0730_plate01, raw_0730_plate01, "0730-plate-01")
plate_0730_02_long <- plate_to_long(layout_0730_plate02, raw_0730_plate02, "0730-plate-02")
plate_0803_01_long <- plate_to_long(layout_0803_plate01, raw_0803_plate01, "0803-plate-01")

cat("--- plate_0730_01_long ---\n")
str(plate_0730_01_long)
cat("\n--- plate_0730_02_long ---\n")
str(plate_0730_02_long)
cat("\n--- plate_0803_01_long ---\n")
str(plate_0803_01_long)
--- plate_0730_01_long ---
'data.frame':   96 obs. of  7 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "0730-plate-01" "0730-plate-01" "0730-plate-01" "0730-plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_02_ambient-glyc-4.7-df.25" ...
 $ luminescence : num  11367 10284 10754 14834 14603 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- plate_0730_02_long ---
'data.frame':   18 obs. of  7 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "0730-plate-02" "0730-plate-02" "0730-plate-02" "0730-plate-02" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "STD-glyc-20" "STD-glyc-20" "STD-glyc-20" "STD-glyc-2" ...
 $ luminescence : num  53166 68390 63329 6947 6432 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- plate_0803_01_long ---
'data.frame':   36 obs. of  7 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "0803-plate-01" "0803-plate-01" "0803-plate-01" "0803-plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "1_06_ambient-glyc-11.0-df.200" ...
 $ luminescence : num  6528 6498 6599 9137 9214 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

5 STANDARD CURVES

Each assay date has its own standard curve (glycogen standards run alongside the samples), fit independently – samples are always back-calculated against the curve from their own assay date/plate, never a curve from a different run.

standards_0730 <- plate_0730_02_long %>%
  filter(grepl("^STD-glyc-", label)) %>%
  mutate(glyc_concentration = as.numeric(sub("^STD-glyc-", "", label)),
         run = "2026-07-30")

negative_0730 <- plate_0730_02_long %>%
  filter(grepl("^NEG-glyc", label)) %>%
  mutate(run = "2026-07-30")

standards_0803 <- plate_0803_01_long %>%
  filter(grepl("^STD-glyc-", label)) %>%
  mutate(glyc_concentration = as.numeric(sub("^STD-glyc-", "", label)),
         run = "2026-08-03")

negative_0803 <- plate_0803_01_long %>%
  filter(grepl("^NEG-glyc", label)) %>%
  mutate(run = "2026-08-03")

cat("--- standards_0730 ---\n")
str(standards_0730)
cat("\n--- standards_0803 ---\n")
str(standards_0803)

cat("\nNegative control, 2026-07-30 (n =", nrow(negative_0730), "): mean luminescence =",
    round(mean(negative_0730$luminescence), 1), "\n")
cat("Negative control, 2026-08-03 (n =", nrow(negative_0803), "): mean luminescence =",
    round(mean(negative_0803$luminescence), 1), "\n")
--- standards_0730 ---
'data.frame':   15 obs. of  9 variables:
 $ plate_row_idx     : int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col         : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate             : chr  "0730-plate-02" "0730-plate-02" "0730-plate-02" "0730-plate-02" ...
 $ plate_row         : chr  "A" "A" "A" "A" ...
 $ well              : chr  "A01" "A02" "A03" "A04" ...
 $ label             : chr  "STD-glyc-20" "STD-glyc-20" "STD-glyc-20" "STD-glyc-2" ...
 $ luminescence      : num  53166 68390 63329 6947 6432 ...
 $ glyc_concentration: num  20 20 20 2 2 2 0.2 0.2 0.2 0.02 ...
 $ run               : chr  "2026-07-30" "2026-07-30" "2026-07-30" "2026-07-30" ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- standards_0803 ---
'data.frame':   18 obs. of  9 variables:
 $ plate_row_idx     : int  3 3 3 3 3 3 3 3 3 3 ...
 $ plate_col         : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate             : chr  "0803-plate-01" "0803-plate-01" "0803-plate-01" "0803-plate-01" ...
 $ plate_row         : chr  "C" "C" "C" "C" ...
 $ well              : chr  "C01" "C02" "C03" "C04" ...
 $ label             : chr  "STD-glyc-20" "STD-glyc-20" "STD-glyc-20" "STD-glyc-10" ...
 $ luminescence      : num  61432 62800 60190 28374 27652 ...
 $ glyc_concentration: num  20 20 20 10 10 10 2 2 2 0.2 ...
 $ run               : chr  "2026-08-03" "2026-08-03" "2026-08-03" "2026-08-03" ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

Negative control, 2026-07-30 (n = 3 ): mean luminescence = 913 
Negative control, 2026-08-03 (n = 3 ): mean luminescence = 583.3 
fit_curve <- function(standards) {
  summ <- standards %>%
    group_by(glyc_concentration) %>%
    summarise(
      glyc_mean_luminescence = mean(luminescence),
      glyc_sd                = sd(luminescence),
      glyc_se                = sd(luminescence) / sqrt(n()),
      glyc_cv                = 100 * sd(luminescence) / mean(luminescence),
      glyc_n                 = n(),
      .groups = "drop"
    ) %>%
    arrange(glyc_concentration)

  lm_model <- lm(glyc_mean_luminescence ~ glyc_concentration, data = summ)

  list(
    summary   = summ,
    model     = lm_model,
    slope     = unname(coef(lm_model)[2]),
    intercept = unname(coef(lm_model)[1]),
    r2        = summary(lm_model)$r.squared,
    conc_min  = min(summ$glyc_concentration[summ$glyc_concentration > 0]),
    conc_max  = max(summ$glyc_concentration),
    fit_min   = min(summ$glyc_concentration),
    fit_max   = max(summ$glyc_concentration)
  )
}

curve_0730 <- fit_curve(standards_0730)
curve_0803 <- fit_curve(standards_0803)

cat("--- 2026-07-30 standard curve ---\n")
str(curve_0730$summary)
cat("slope:", curve_0730$slope, " intercept:", curve_0730$intercept,
    " R2:", round(curve_0730$r2, 5), "\n")
cat("quantifiable range:", curve_0730$conc_min, "-", curve_0730$conc_max, "ug/mL\n\n")

cat("--- 2026-08-03 standard curve ---\n")
str(curve_0803$summary)
cat("slope:", curve_0803$slope, " intercept:", curve_0803$intercept,
    " R2:", round(curve_0803$r2, 5), "\n")
cat("quantifiable range:", curve_0803$conc_min, "-", curve_0803$conc_max, "ug/mL\n")
--- 2026-07-30 standard curve ---
tibble [5 × 6] (S3: tbl_df/tbl/data.frame)
 $ glyc_concentration    : num [1:5] 0 0.02 0.2 2 20
 $ glyc_mean_luminescence: num [1:5] 910 1087 1410 6512 61628
 $ glyc_sd               : num [1:5] 13.5 107.2 30.4 401.5 7753.2
 $ glyc_se               : num [1:5] 7.81 61.89 17.52 231.79 4476.3
 $ glyc_cv               : num [1:5] 1.49 9.86 2.15 6.17 12.58
 $ glyc_n                : int [1:5] 3 3 3 3 3
slope: 3039.495  intercept: 801.7511  R2: 0.99993 
quantifiable range: 0.02 - 20 ug/mL

--- 2026-08-03 standard curve ---
tibble [6 × 6] (S3: tbl_df/tbl/data.frame)
 $ glyc_concentration    : num [1:6] 0 0.02 0.2 2 10 20
 $ glyc_mean_luminescence: num [1:6] 313 681 2220 12126 28236 ...
 $ glyc_sd               : num [1:6] 256 107 1100 338 529 ...
 $ glyc_se               : num [1:6] 147.7 61.9 635.4 194.9 305.5 ...
 $ glyc_cv               : num [1:6] 81.82 15.73 49.57 2.78 1.87 ...
 $ glyc_n                : int [1:6] 3 3 3 3 3 3
slope: 2937.696  intercept: 1732.959  R2: 0.98861 
quantifiable range: 0.02 - 20 ug/mL

5.1 Plot standard curves

curve_plot_data <- bind_rows(
  curve_0730$summary %>% mutate(run = "2026-07-30"),
  curve_0803$summary %>% mutate(run = "2026-08-03")
)
curve_fit_lines <- bind_rows(
  data.frame(run = "2026-07-30", slope = curve_0730$slope, intercept = curve_0730$intercept,
             fit_min = curve_0730$fit_min, fit_max = curve_0730$fit_max, r2 = curve_0730$r2),
  data.frame(run = "2026-08-03", slope = curve_0803$slope, intercept = curve_0803$intercept,
             fit_min = curve_0803$fit_min, fit_max = curve_0803$fit_max, r2 = curve_0803$r2)
) %>%
  mutate(
    y_start   = intercept + slope * fit_min,
    y_end     = intercept + slope * fit_max,
    r2_label  = paste0("R\u00b2 = ", sprintf("%.4f", r2))
  )

cat("--- curve_fit_lines: trend-line endpoints + R2 per run ---\n")
str(curve_fit_lines)

standard_curve_plot <- ggplot(curve_plot_data, aes(x = glyc_concentration, y = glyc_mean_luminescence)) +
  geom_segment(data = curve_fit_lines,
               aes(x = fit_min, xend = fit_max, y = y_start, yend = y_end),
               color = "steelblue", linewidth = 0.6, inherit.aes = FALSE) +
  geom_errorbar(aes(ymin = glyc_mean_luminescence - glyc_se, ymax = glyc_mean_luminescence + glyc_se),
                width = 0.3, color = "grey40") +
  geom_point(size = 2.4, color = "steelblue") +
  geom_text(data = curve_fit_lines, aes(x = -Inf, y = Inf, label = r2_label),
            hjust = -0.15, vjust = 1.8, inherit.aes = FALSE, size = 3.6, fontface = "bold") +
  facet_wrap(~ run, scales = "free_y") +
  labs(x = "Glycogen standard concentration (\u00b5g/mL)",
       y = "Mean luminescence (RLU)",
       title = "Standard curves by assay date") +
  theme_bw(base_size = 12) +
  theme(strip.text = element_text(face = "bold"))

print(standard_curve_plot)

ggsave(file.path(output_dir, "standard_curves_both_runs.png"), standard_curve_plot,
       width = 10, height = 5, dpi = 300)
--- curve_fit_lines: trend-line endpoints + R2 per run ---
'data.frame':   2 obs. of  9 variables:
 $ run      : chr  "2026-07-30" "2026-08-03"
 $ slope    : num  3039 2938
 $ intercept: num  802 1733
 $ fit_min  : num  0 0
 $ fit_max  : num  20 20
 $ r2       : num  1 0.989
 $ y_start  : num  802 1733
 $ y_end    : num  61592 60487
 $ r2_label : chr  "R² = 0.9999" "R² = 0.9886"

6 SAMPLE LABEL PARSING

Sample labels are parsed into family, individual, temperature, tissue weight, and dilution factor. Both runs’ sample wells (excluding standards and the negative control) use the same label grammar.

parse_samples <- function(plate_long) {
  plate_long %>%
    filter(!grepl("^STD-glyc-|^NEG-glyc", label)) %>%
    mutate(label_clean = sub("^([19])[-_]([0-9]{2})[-_](ambient|36C)-", "\\1_\\2_\\3-", label)) %>%
    mutate(
      sample_id   = sub("-glyc-.*$", "", label_clean),
      family      = sub("^([19])_.*$", "\\1", label_clean),
      individual  = sub("^[19]_([0-9]{2})_.*$", "\\1", label_clean),
      temperature = sub("^[19]_[0-9]{2}_([^-]+)-.*$", "\\1", label_clean),
      weight_mg   = as.numeric(sub("^.*-glyc-([0-9.]+)-df\\..*$", "\\1", label_clean)),
      dilution    = as.numeric(sub("^.*-df\\.", "", label_clean))
    ) %>%
    mutate(
      family      = factor(paste("Family", family), levels = c("Family 1", "Family 9")),
      temperature = factor(temperature, levels = c("ambient", "36C"))
    )
}
samples_0730 <- parse_samples(plate_0730_01_long)
samples_0803 <- parse_samples(plate_0803_01_long)

cat("--- samples_0730 ---\n")
str(samples_0730)
cat("\n--- samples_0803 ---\n")
str(samples_0803)

cat("\n2026-07-30: ", length(unique(samples_0730$sample_id)), "unique samples,",
    nrow(samples_0730), "wells (triplicate).\n")
cat("2026-08-03: ", length(unique(samples_0803$sample_id)), "unique samples,",
    nrow(samples_0803), "wells (triplicate).\n")
--- samples_0730 ---
'data.frame':   96 obs. of  14 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "0730-plate-01" "0730-plate-01" "0730-plate-01" "0730-plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_02_ambient-glyc-4.7-df.25" ...
 $ luminescence : num  11367 10284 10754 14834 14603 ...
 $ label_clean  : chr  "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_02_ambient-glyc-4.7-df.25" ...
 $ sample_id    : chr  "1_01_ambient" "1_01_ambient" "1_01_ambient" "1_02_ambient" ...
 $ family       : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual   : chr  "01" "01" "01" "02" ...
 $ temperature  : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 1 1 ...
 $ weight_mg    : num  7 7 7 4.7 4.7 4.7 2.5 2.5 2.5 10.7 ...
 $ dilution     : num  25 25 25 25 25 25 25 25 25 25 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- samples_0803 ---
'data.frame':   15 obs. of  14 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "0803-plate-01" "0803-plate-01" "0803-plate-01" "0803-plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "1_06_ambient-glyc-11.0-df.200" ...
 $ luminescence : num  6528 6498 6599 9137 9214 ...
 $ label_clean  : chr  "9_05_ambient-glyc-9.2-df.200" "9_05_ambient-glyc-9.2-df.200" "9_05_ambient-glyc-9.2-df.200" "1_06_ambient-glyc-11.0-df.200" ...
 $ sample_id    : chr  "9_05_ambient" "9_05_ambient" "9_05_ambient" "1_06_ambient" ...
 $ family       : Factor w/ 2 levels "Family 1","Family 9": 2 2 2 1 1 1 1 1 1 1 ...
 $ individual   : chr  "05" "05" "05" "06" ...
 $ temperature  : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 1 2 ...
 $ weight_mg    : num  9.2 9.2 9.2 11 11 11 24.8 24.8 24.8 12.4 ...
 $ dilution     : num  200 200 200 200 200 200 200 200 200 200 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

2026-07-30:  32 unique samples, 96 wells (triplicate).
2026-08-03:  5 unique samples, 15 wells (triplicate).

7 PER-SAMPLE GLYCOGEN QUANTIFICATION

For each sample, technical-replicate wells are averaged, back-calculated against that assay date’s own standard curve, corrected for dilution factor and tissue weight, and flagged for whether the well-level luminescence falls within the quantifiable range of that date’s standard curve.

calc_glycogen <- function(samples_wells, curve, run_label) {
  samples_wells %>%
    group_by(sample_id, family, individual, temperature, weight_mg, dilution) %>%
    summarise(
      n_reps   = n(),
      mean_lum = mean(luminescence),
      sd_lum   = sd(luminescence),
      cv_lum   = 100 * sd(luminescence) / mean(luminescence),
      .groups  = "drop"
    ) %>%
    mutate(
      run                   = run_label,
      well_conc_ug_mL       = (mean_lum - curve$intercept) / curve$slope,
      homogenate_conc_ug_mL = well_conc_ug_mL * dilution,
      norm_glycogen         = homogenate_conc_ug_mL / weight_mg,
      in_std_range          = well_conc_ug_mL >= curve$conc_min & well_conc_ug_mL <= curve$conc_max
    ) %>%
    arrange(family, temperature, individual)
}
glycogen_0730 <- calc_glycogen(samples_0730, curve_0730, "2026-07-30")
glycogen_0803 <- calc_glycogen(samples_0803, curve_0803, "2026-08-03")

cat("--- glycogen_0730 ---\n")
str(glycogen_0730)
cat("\n--- glycogen_0803 ---\n")
str(glycogen_0803)
--- glycogen_0730 ---
tibble [32 × 15] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:32] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:32] "01" "02" "03" "04" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 2 2 ...
 $ weight_mg            : num [1:32] 7 4.7 2.5 10.7 7.9 11 6.1 24.8 9.6 9.2 ...
 $ dilution             : num [1:32] 25 25 25 25 25 25 25 25 25 25 ...
 $ n_reps               : int [1:32] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:32] 10802 14632 7373 39486 22099 ...
 $ sd_lum               : num [1:32] 543 189 278 2525 603 ...
 $ cv_lum               : num [1:32] 5.03 1.29 3.77 6.4 2.73 ...
 $ run                  : chr [1:32] "2026-07-30" "2026-07-30" "2026-07-30" "2026-07-30" ...
 $ well_conc_ug_mL      : num [1:32] 3.29 4.55 2.16 12.73 7.01 ...
 $ homogenate_conc_ug_mL: num [1:32] 82.2 113.8 54 318.2 175.2 ...
 $ norm_glycogen        : num [1:32] 11.7 24.2 21.6 29.7 22.2 ...
 $ in_std_range         : logi [1:32] TRUE TRUE TRUE TRUE TRUE FALSE ...

--- glycogen_0803 ---
tibble [5 × 15] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:5] "1_06_ambient" "1_08_ambient" "1_04_36C" "1_05_36C" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 2
 $ individual           : chr [1:5] "06" "08" "04" "05" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 2 2 1
 $ weight_mg            : num [1:5] 11 24.8 12.4 25.7 9.2
 $ dilution             : num [1:5] 200 200 200 200 200
 $ n_reps               : int [1:5] 3 3 3 3 3
 $ mean_lum             : num [1:5] 9218 30767 823 8742 6542
 $ sd_lum               : num [1:5] 83.6 278 42.9 236.4 51.9
 $ cv_lum               : num [1:5] 0.907 0.903 5.211 2.704 0.793
 $ run                  : chr [1:5] "2026-08-03" "2026-08-03" "2026-08-03" "2026-08-03" ...
 $ well_conc_ug_mL      : num [1:5] 2.55 9.88 -0.31 2.39 1.64
 $ homogenate_conc_ug_mL: num [1:5] 510 1977 -62 477 327
 $ norm_glycogen        : num [1:5] 46.3 79.7 -5 18.6 35.6
 $ in_std_range         : logi [1:5] TRUE TRUE FALSE TRUE TRUE

8 COEFFICIENT OF VARIATION (CV) CHECK

Technical-replicate CV is computed for every sample in both runs. Samples exceeding 15% CV are flagged, not excluded – per the analysis requirements, high CV is a data-quality note carried alongside the result, not a reason to drop a sample.

all_glycogen_by_run <- bind_rows(glycogen_0730, glycogen_0803) %>%
  mutate(high_cv = cv_lum > 15)

cat("--- all_glycogen_by_run: every measurement from both runs, CV-flagged ---\n")
str(all_glycogen_by_run)

high_cv_samples <- all_glycogen_by_run %>%
  filter(high_cv) %>%
  select(sample_id, run, n_reps, mean_lum, cv_lum, norm_glycogen, in_std_range)

cat("\nSamples exceeding 15% technical-replicate CV (flagged, not excluded):\n")
kable(high_cv_samples, digits = 2)
--- all_glycogen_by_run: every measurement from both runs, CV-flagged ---
tibble [37 × 16] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:37] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:37] "01" "02" "03" "04" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 2 2 ...
 $ weight_mg            : num [1:37] 7 4.7 2.5 10.7 7.9 11 6.1 24.8 9.6 9.2 ...
 $ dilution             : num [1:37] 25 25 25 25 25 25 25 25 25 25 ...
 $ n_reps               : int [1:37] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:37] 10802 14632 7373 39486 22099 ...
 $ sd_lum               : num [1:37] 543 189 278 2525 603 ...
 $ cv_lum               : num [1:37] 5.03 1.29 3.77 6.4 2.73 ...
 $ run                  : chr [1:37] "2026-07-30" "2026-07-30" "2026-07-30" "2026-07-30" ...
 $ well_conc_ug_mL      : num [1:37] 3.29 4.55 2.16 12.73 7.01 ...
 $ homogenate_conc_ug_mL: num [1:37] 82.2 113.8 54 318.2 175.2 ...
 $ norm_glycogen        : num [1:37] 11.7 24.2 21.6 29.7 22.2 ...
 $ in_std_range         : logi [1:37] TRUE TRUE TRUE TRUE TRUE FALSE ...
 $ high_cv              : logi [1:37] FALSE FALSE FALSE FALSE FALSE FALSE ...

Samples exceeding 15% technical-replicate CV (flagged, not excluded):
sample_id run n_reps mean_lum cv_lum norm_glycogen in_std_range
1_04_36C 2026-07-30 3 5075.33 30.07 2.83 TRUE

9 STANDARD CURVES WITH SAMPLES OVERLAID

Each run’s samples are plotted against that run’s own standard curve, on the well-concentration scale, so it’s visually clear which samples fall inside vs. outside the curve’s quantifiable range.

9.1 Plot 20260730 Standard Curves with Samples

r2_label_0730 <- paste0("R\u00b2 = ", sprintf("%.4f", curve_0730$r2))

std_samples_0730 <- ggplot() +
  geom_ribbon(data = data.frame(x = c(curve_0730$conc_min, curve_0730$conc_max)),
              aes(x = x, ymin = -Inf, ymax = Inf), fill = "grey85", alpha = 0.4,
              inherit.aes = FALSE) +
  geom_segment(aes(x = curve_0730$fit_min, xend = curve_0730$fit_max,
                    y = curve_0730$intercept + curve_0730$slope * curve_0730$fit_min,
                    yend = curve_0730$intercept + curve_0730$slope * curve_0730$fit_max),
               color = "steelblue", linewidth = 0.6) +
  geom_point(data = curve_0730$summary, aes(x = glyc_concentration, y = glyc_mean_luminescence),
             size = 2.4, color = "steelblue") +
  geom_point(data = glycogen_0730, aes(x = well_conc_ug_mL, y = mean_lum, shape = in_std_range),
             size = 2.2, color = "firebrick", alpha = 0.75) +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                      labels = c(`TRUE` = "in range", `FALSE` = "out of range"),
                      name = "Sample status") +
  annotate("text", x = -Inf, y = Inf, label = r2_label_0730,
           hjust = -0.15, vjust = 1.8, size = 3.6, fontface = "bold") +
  labs(x = "Glycogen concentration (\u00b5g/mL, well-level)", y = "Luminescence (RLU)",
       title = "2026-07-30: standard curve with samples overlaid",
       subtitle = "Shaded band = quantifiable range of the standard curve") +
  theme_bw(base_size = 12)

print(std_samples_0730)

ggsave(file.path(output_dir, "standard_curve_with_samples_20260730.png"), std_samples_0730,
       width = 8, height = 5.5, dpi = 300)

9.2 Plot 20260803 Standard Curves with Samples

r2_label_0803 <- paste0("R\u00b2 = ", sprintf("%.4f", curve_0803$r2))

std_samples_0803 <- ggplot() +
  geom_ribbon(data = data.frame(x = c(curve_0803$conc_min, curve_0803$conc_max)),
              aes(x = x, ymin = -Inf, ymax = Inf), fill = "grey85", alpha = 0.4,
              inherit.aes = FALSE) +
  geom_segment(aes(x = curve_0803$fit_min, xend = curve_0803$fit_max,
                    y = curve_0803$intercept + curve_0803$slope * curve_0803$fit_min,
                    yend = curve_0803$intercept + curve_0803$slope * curve_0803$fit_max),
               color = "steelblue", linewidth = 0.6) +
  geom_point(data = curve_0803$summary, aes(x = glyc_concentration, y = glyc_mean_luminescence),
             size = 2.4, color = "steelblue") +
  geom_point(data = glycogen_0803, aes(x = well_conc_ug_mL, y = mean_lum, shape = in_std_range),
             size = 2.2, color = "firebrick", alpha = 0.75) +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                      labels = c(`TRUE` = "in range", `FALSE` = "out of range"),
                      name = "Sample status") +
  annotate("text", x = -Inf, y = Inf, label = r2_label_0803,
           hjust = -0.15, vjust = 1.8, size = 3.6, fontface = "bold") +
  labs(x = "Glycogen concentration (\u00b5g/mL, well-level)", y = "Luminescence (RLU)",
       title = "2026-08-03: standard curve with re-assayed samples overlaid",
       subtitle = "Shaded band = quantifiable range of the standard curve") +
  theme_bw(base_size = 12)

print(std_samples_0803)

ggsave(file.path(output_dir, "standard_curve_with_samples_20260803.png"), std_samples_0803,
       width = 8, height = 5.5, dpi = 300)

10 MERGING RUNS INTO A SINGLE FINAL DATASET

For the 5 samples with a measurement from both dates, the value used going forward is the one that falls within its own run’s standard-curve range (interpolated), with the more recent run used as a tiebreaker if both qualify. All other 27 samples have only a single (2026-07-30) measurement and pass through unchanged. This yields the reconciliation described in Background: the 4 originally-extrapolated samples take their 2026-08-03 (1:200) value, while 1_04_36C takes its 2026-07-30 (1:25) value, since its 2026-08-03 re-assay fell below the standard curve’s floor (negative back-calculated concentration).

combined_measurements <- bind_rows(glycogen_0730, glycogen_0803) %>%
  mutate(high_cv = cv_lum > 15) %>%
  arrange(sample_id, run)

cat("--- combined_measurements: every available measurement, both runs, before reconciliation ---\n")
str(combined_measurements)

final_glycogen <- combined_measurements %>%
  group_by(sample_id) %>%
  arrange(desc(in_std_range), desc(run), .by_group = TRUE) %>%
  mutate(n_runs_available = n()) %>%
  slice(1) %>%
  ungroup() %>%
  arrange(family, temperature, individual)

cat("\n--- final_glycogen: one reconciled measurement per sample ---\n")
str(final_glycogen)

cat("\nSamples for which the 2026-08-03 re-assay superseded the 2026-07-30 value:\n")
print(as.data.frame(final_glycogen %>%
  filter(n_runs_available > 1, run == "2026-08-03") %>%
  select(sample_id, run, in_std_range, cv_lum, norm_glycogen)))

cat("\nSamples for which the 2026-07-30 value was retained despite a 2026-08-03 re-assay existing:\n")
print(as.data.frame(final_glycogen %>%
  filter(n_runs_available > 1, run == "2026-07-30") %>%
  select(sample_id, run, in_std_range, cv_lum, norm_glycogen)))

cat("\nTotal samples in final dataset:", nrow(final_glycogen), "\n")
cat("Samples still out-of-range after reconciliation:",
    sum(!final_glycogen$in_std_range), "\n")
--- combined_measurements: every available measurement, both runs, before reconciliation ---
tibble [37 × 16] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:37] "1_01_36C" "1_01_ambient" "1_02_36C" "1_02_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:37] "01" "01" "02" "02" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 2 1 2 1 2 1 2 2 1 2 ...
 $ weight_mg            : num [1:37] 9.6 7 9.2 4.7 3.1 2.5 12.4 12.4 10.7 25.7 ...
 $ dilution             : num [1:37] 25 25 25 25 25 25 25 200 25 25 ...
 $ n_reps               : int [1:37] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:37] 31844 10802 5679 14632 4832 ...
 $ sd_lum               : num [1:37] 1604 543 377 189 479 ...
 $ cv_lum               : num [1:37] 5.04 5.03 6.63 1.29 9.92 ...
 $ run                  : chr [1:37] "2026-07-30" "2026-07-30" "2026-07-30" "2026-07-30" ...
 $ well_conc_ug_mL      : num [1:37] 10.21 3.29 1.6 4.55 1.33 ...
 $ homogenate_conc_ug_mL: num [1:37] 255.3 82.2 40.1 113.8 33.1 ...
 $ norm_glycogen        : num [1:37] 26.6 11.75 4.36 24.2 10.69 ...
 $ in_std_range         : logi [1:37] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ high_cv              : logi [1:37] FALSE FALSE FALSE FALSE FALSE FALSE ...

--- final_glycogen: one reconciled measurement per sample ---
tibble [32 × 17] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:32] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:32] "01" "02" "03" "04" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 2 2 ...
 $ weight_mg            : num [1:32] 7 4.7 2.5 10.7 7.9 11 6.1 24.8 9.6 9.2 ...
 $ dilution             : num [1:32] 25 25 25 25 25 200 25 200 25 25 ...
 $ n_reps               : int [1:32] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:32] 10802 14632 7373 39486 22099 ...
 $ sd_lum               : num [1:32] 543 189 278 2525 603 ...
 $ cv_lum               : num [1:32] 5.03 1.29 3.77 6.4 2.73 ...
 $ run                  : chr [1:32] "2026-07-30" "2026-07-30" "2026-07-30" "2026-07-30" ...
 $ well_conc_ug_mL      : num [1:32] 3.29 4.55 2.16 12.73 7.01 ...
 $ homogenate_conc_ug_mL: num [1:32] 82.2 113.8 54 318.2 175.2 ...
 $ norm_glycogen        : num [1:32] 11.7 24.2 21.6 29.7 22.2 ...
 $ in_std_range         : logi [1:32] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ high_cv              : logi [1:32] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_runs_available     : int [1:32] 1 1 1 1 1 2 1 2 1 1 ...

Samples for which the 2026-08-03 re-assay superseded the 2026-07-30 value:
     sample_id        run in_std_range    cv_lum norm_glycogen
1 1_06_ambient 2026-08-03         TRUE 0.9067180      46.32804
2 1_08_ambient 2026-08-03         TRUE 0.9034992      79.70469
3     1_05_36C 2026-08-03         TRUE 2.7043812      18.56729
4 9_05_ambient 2026-08-03         TRUE 0.7928931      35.58473

Samples for which the 2026-07-30 value was retained despite a 2026-08-03 re-assay existing:
  sample_id        run in_std_range  cv_lum norm_glycogen
1  1_04_36C 2026-07-30         TRUE 30.0656      2.834712

Total samples in final dataset: 32 
Samples still out-of-range after reconciliation: 0

10.1 Excluding out-of-range samples from statistical comparisons

Per the analysis requirements, any sample still falling outside its standard curve’s quantifiable range after reconciliation is excluded from the statistical comparisons below (family/temperature ANOVA and pairwise tests), though it remains visible in the full results table and raw distribution plots.

stats_data <- final_glycogen %>% filter(in_std_range)

cat("Samples excluded from statistical comparisons (out of range after reconciliation):\n")
print(as.data.frame(final_glycogen %>% filter(!in_std_range) %>%
  select(sample_id, run, in_std_range, norm_glycogen)))

cat("\nn samples entering statistical comparisons:", nrow(stats_data), "of", nrow(final_glycogen), "\n")
str(stats_data)
Samples excluded from statistical comparisons (out of range after reconciliation):
[1] sample_id     run           in_std_range  norm_glycogen
<0 rows> (or 0-length row.names)

n samples entering statistical comparisons: 32 of 32 
tibble [32 × 17] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:32] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:32] "01" "02" "03" "04" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 2 2 ...
 $ weight_mg            : num [1:32] 7 4.7 2.5 10.7 7.9 11 6.1 24.8 9.6 9.2 ...
 $ dilution             : num [1:32] 25 25 25 25 25 200 25 200 25 25 ...
 $ n_reps               : int [1:32] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:32] 10802 14632 7373 39486 22099 ...
 $ sd_lum               : num [1:32] 543 189 278 2525 603 ...
 $ cv_lum               : num [1:32] 5.03 1.29 3.77 6.4 2.73 ...
 $ run                  : chr [1:32] "2026-07-30" "2026-07-30" "2026-07-30" "2026-07-30" ...
 $ well_conc_ug_mL      : num [1:32] 3.29 4.55 2.16 12.73 7.01 ...
 $ homogenate_conc_ug_mL: num [1:32] 82.2 113.8 54 318.2 175.2 ...
 $ norm_glycogen        : num [1:32] 11.7 24.2 21.6 29.7 22.2 ...
 $ in_std_range         : logi [1:32] TRUE TRUE TRUE TRUE TRUE TRUE ...
 $ high_cv              : logi [1:32] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ n_runs_available     : int [1:32] 1 1 1 1 1 2 1 2 1 1 ...

11 RESULTS TABLE

results_table <- final_glycogen %>%
  transmute(
    Sample = sample_id,
    Family = family,
    Temperature = temperature,
    Individual = individual,
    `Tissue (mg)` = weight_mg,
    Dilution = dilution,
    `Run used` = run,
    `Mean lum.` = round(mean_lum, 1),
    `CV (%)` = round(cv_lum, 1),
    `High CV (>15%)` = ifelse(high_cv, "yes", "no"),
    `Well glycogen (ug/mL)` = round(well_conc_ug_mL, 3),
    `Normalized glycogen (ug/mL/mg)` = round(norm_glycogen, 2),
    `In std range` = ifelse(in_std_range, "yes", "NO - excluded from stats")
  )

kable(results_table)

write.csv(results_table, file.path(output_dir, "sample_glycogen_results_combined.csv"), row.names = FALSE)
Sample Family Temperature Individual Tissue (mg) Dilution Run used Mean lum. CV (%) High CV (>15%) Well glycogen (ug/mL) Normalized glycogen (ug/mL/mg) In std range
1_01_ambient Family 1 ambient 01 7.0 25 2026-07-30 10801.7 5.0 no 3.290 11.75 yes
1_02_ambient Family 1 ambient 02 4.7 25 2026-07-30 14632.3 1.3 no 4.550 24.20 yes
1_03_ambient Family 1 ambient 03 2.5 25 2026-07-30 7372.7 3.8 no 2.162 21.62 yes
1_04_ambient Family 1 ambient 04 10.7 25 2026-07-30 39485.7 6.4 no 12.727 29.74 yes
1_05_ambient Family 1 ambient 05 7.9 25 2026-07-30 22099.3 2.7 no 7.007 22.17 yes
1_06_ambient Family 1 ambient 06 11.0 200 2026-08-03 9218.3 0.9 no 2.548 46.33 yes
1_07_ambient Family 1 ambient 07 6.1 25 2026-07-30 5075.7 5.4 no 1.406 5.76 yes
1_08_ambient Family 1 ambient 08 24.8 200 2026-08-03 30767.3 0.9 no 9.883 79.70 yes
1_01_36C Family 1 36C 01 9.6 25 2026-07-30 31844.0 5.0 no 10.213 26.60 yes
1_02_36C Family 1 36C 02 9.2 25 2026-07-30 5679.0 6.6 no 1.605 4.36 yes
1_03_36C Family 1 36C 03 3.1 25 2026-07-30 4831.7 9.9 no 1.326 10.69 yes
1_04_36C Family 1 36C 04 12.4 25 2026-07-30 5075.3 30.1 yes 1.406 2.83 yes
1_05_36C Family 1 36C 05 25.7 200 2026-08-03 8742.0 2.7 no 2.386 18.57 yes
1_06_36C Family 1 36C 06 13.0 25 2026-07-30 56800.0 1.9 no 18.424 35.43 yes
1_07_36C Family 1 36C 07 6.1 25 2026-07-30 21659.7 4.1 no 6.862 28.12 yes
1_08_36C Family 1 36C 08 8.8 25 2026-07-30 46235.0 3.5 no 14.948 42.46 yes
9_01_ambient Family 9 ambient 01 7.3 25 2026-07-30 7336.3 3.3 no 2.150 7.36 yes
9_02_ambient Family 9 ambient 02 4.4 25 2026-07-30 4990.7 4.7 no 1.378 7.83 yes
9_03_ambient Family 9 ambient 03 4.3 25 2026-07-30 11467.0 5.1 no 3.509 20.40 yes
9_04_ambient Family 9 ambient 04 1.8 25 2026-07-30 5652.7 6.4 no 1.596 22.17 yes
9_05_ambient Family 9 ambient 05 9.2 200 2026-08-03 6541.7 0.8 no 1.637 35.58 yes
9_06_ambient Family 9 ambient 06 4.3 25 2026-07-30 3648.7 6.5 no 0.937 5.45 yes
9_07_ambient Family 9 ambient 07 2.4 25 2026-07-30 9969.0 4.9 no 3.016 31.42 yes
9_08_ambient Family 9 ambient 08 7.4 25 2026-07-30 2677.7 5.3 no 0.617 2.09 yes
9_01_36C Family 9 36C 01 7.7 25 2026-07-30 14412.7 5.9 no 4.478 14.54 yes
9_02_36C Family 9 36C 02 15.7 25 2026-07-30 12523.3 4.4 no 3.856 6.14 yes
9_03_36C Family 9 36C 03 10.4 25 2026-07-30 43856.7 3.2 no 14.165 34.05 yes
9_04_36C Family 9 36C 04 11.3 25 2026-07-30 23983.3 2.2 no 7.627 16.87 yes
9_05_36C Family 9 36C 05 9.2 25 2026-07-30 49780.7 4.3 no 16.114 43.79 yes
9_06_36C Family 9 36C 06 4.3 25 2026-07-30 14106.0 3.2 no 4.377 25.45 yes
9_07_36C Family 9 36C 07 2.4 25 2026-07-30 2408.0 11.0 no 0.528 5.50 yes
9_08_36C Family 9 36C 08 4.9 25 2026-07-30 18651.0 6.6 no 5.872 29.96 yes

12 GROUP SUMMARY STATISTICS

group_summary_stats <- stats_data %>%
  group_by(Family = family, Temperature = temperature) %>%
  summarise(
    n      = n(),
    mean   = mean(norm_glycogen),
    sd     = sd(norm_glycogen),
    se     = sd(norm_glycogen) / sqrt(n()),
    median = median(norm_glycogen),
    .groups = "drop"
  )

cat("--- group_summary_stats ---\n")
str(group_summary_stats)
kable(group_summary_stats, digits = 2)

write.csv(group_summary_stats, file.path(output_dir, "group_summary_stats_combined.csv"), row.names = FALSE)
--- group_summary_stats ---
tibble [4 × 7] (S3: tbl_df/tbl/data.frame)
 $ Family     : Factor w/ 2 levels "Family 1","Family 9": 1 1 2 2
 $ Temperature: Factor w/ 2 levels "ambient","36C": 1 2 1 2
 $ n          : int [1:4] 8 8 8 8
 $ mean       : num [1:4] 30.2 21.1 16.5 22
 $ sd         : num [1:4] 23.4 14.5 12.7 13.6
 $ se         : num [1:4] 8.26 5.12 4.48 4.82
 $ median     : num [1:4] 23.2 22.6 14.1 21.2
Family Temperature n mean sd se median
Family 1 ambient 8 30.16 23.37 8.26 23.19
Family 1 36C 8 21.13 14.49 5.12 22.58
Family 9 ambient 8 16.54 12.66 4.48 14.12
Family 9 36C 8 22.04 13.63 4.82 21.16

13 STATISTICAL ANALYSIS

Statistical analysis is performed before the box plot and interaction plot below, so that the plots can be annotated with the significance calls made here. Normalized glycogen values span roughly an order of magnitude and are right-skewed, so the primary test uses a log10 transform to better meet ANOVA’s normality/homoscedasticity assumptions; the raw-scale ANOVA and a non-parametric (Wilcoxon rank-sum) alternative are reported alongside it as robustness checks. Out-of-range samples (if any remain after reconciliation) are excluded from all tests below (stats_data, 32 samples).

13.1 Two-way ANOVA (family x temperature)

aov_log <- aov(log10(norm_glycogen) ~ family * temperature, data = stats_data)
cat("--- Two-way ANOVA, log10(normalized glycogen) ---\n")
print(summary(aov_log))
--- Two-way ANOVA, log10(normalized glycogen) ---
                   Df Sum Sq Mean Sq F value Pr(>F)
family              1  0.111 0.11078   0.730  0.400
temperature         1  0.000 0.00005   0.000  0.986
family:temperature  1  0.272 0.27183   1.791  0.192
Residuals          28  4.249 0.15176               
aov_raw <- aov(norm_glycogen ~ family * temperature, data = stats_data)
cat("--- Two-way ANOVA, raw-scale normalized glycogen (robustness check) ---\n")
print(summary(aov_raw))
--- Two-way ANOVA, raw-scale normalized glycogen (robustness check) ---
                   Df Sum Sq Mean Sq F value Pr(>F)
family              1    324   323.5   1.174  0.288
temperature         1     25    24.8   0.090  0.766
family:temperature  1    422   422.1   1.532  0.226
Residuals          28   7716   275.6

13.2 Overall family and temperature comparisons (Wilcoxon rank-sum)

wilcox_family <- wilcox.test(norm_glycogen ~ family, data = stats_data)
wilcox_temp   <- wilcox.test(norm_glycogen ~ temperature, data = stats_data)

cat("Family 1 vs. Family 9 (pooled across temperature):\n")
print(wilcox_family)
cat("\nAmbient vs. 36C (pooled across family):\n")
print(wilcox_temp)
Family 1 vs. Family 9 (pooled across temperature):

    Wilcoxon rank sum exact test

data:  norm_glycogen by family
W = 148, p-value = 0.4677
alternative hypothesis: true location shift is not equal to 0


Ambient vs. 36C (pooled across family):

    Wilcoxon rank sum exact test

data:  norm_glycogen by temperature
W = 127, p-value = 0.9852
alternative hypothesis: true location shift is not equal to 0

13.3 Reusable significance-annotation helpers

These three helpers are plot-agnostic: they take a family/temperature-style pairwise-comparison table (built in the next section) and a data frame, and compute/draw asterisk-annotated brackets. They generalize directly to other analyses (different family counts, additional temperature levels, or different grouping factors) as long as the same long-format comparison table shape (comparison_type, fixed_level, group1, group2, p_value) is provided. Defined here, ahead of the comparisons themselves, since the comparison-building chunk below calls sig_stars() directly.

# sig_stars(): map p-values to conventional significance stars.
# Returns NA (i.e. "no annotation") for non-significant results, so callers
# can filter on !is.na(sig) to find only what needs to be drawn.
sig_stars <- function(p, thresholds = c(0.001, 0.01, 0.05), symbols = c("***", "**", "*")) {
  ord <- order(thresholds)
  thresholds <- thresholds[ord]; symbols <- symbols[ord]
  out <- rep(NA_character_, length(p))
  for (i in seq_along(thresholds)) out[is.na(out) & p < thresholds[i]] <- symbols[i]
  out
}

# dodge_offset(): x-axis offset of the k-th level of an n-level dodged group,
# matching ggplot2's position_dodge(width)/position_jitterdodge(dodge.width = width).
dodge_offset <- function(group_index, n_groups, width = 0.75) {
  -width / 2 + width / n_groups * (group_index - 0.5)
}

# build_sig_brackets(): compute bracket endpoints + stacked heights for every
# significant (non-NA sig) row of `comparisons`, for a two-factor plot with
# `x_var` on the x-axis and `dodge_var` as the fill/color/dodge factor.
#   - comparison_type == "across_x"     : fixed dodge_var level, compare two
#                                         x_var levels (horizontal bracket
#                                         spanning x positions).
#   - comparison_type == "across_dodge" : fixed x_var level, compare two
#                                         dodge_var levels (horizontal bracket
#                                         spanning dodge offsets within one
#                                         x position; dodge_width = 0 collapses
#                                         this to a zero-width case, so only
#                                         use "across_dodge" when dodge_width > 0).
# Returns list(brackets, tip) where brackets is empty (0 rows) when nothing
# is significant -- callers should add zero layers in that case, which
# geom_sig_brackets() does automatically.
build_sig_brackets <- function(data, y_var, x_var, dodge_var, comparisons,
                                dodge_width = 0.75, step = NULL, tip = NULL) {
  y_vals  <- data[[y_var]]
  y_range <- diff(range(y_vals, na.rm = TRUE))
  if (is.null(step)) step <- 0.09 * y_range
  if (is.null(tip))  tip  <- 0.02 * y_range

  x_levels     <- levels(data[[x_var]])
  dodge_levels <- levels(data[[dodge_var]])
  n_dodge      <- length(dodge_levels)

  offset_for <- function(lvl) dodge_offset(match(lvl, dodge_levels), n_dodge, dodge_width)
  xnum_for   <- function(lvl) match(lvl, x_levels)

  sig_rows <- comparisons %>% filter(!is.na(sig))
  if (nrow(sig_rows) == 0) return(list(brackets = sig_rows, tip = tip))

  base_y <- max(y_vals, na.rm = TRUE) + step

  brackets <- sig_rows %>%
    rowwise() %>%
    mutate(
      x1 = if (comparison_type == "across_x") xnum_for(group1) + offset_for(fixed_level)
           else                               xnum_for(fixed_level) + offset_for(group1),
      x2 = if (comparison_type == "across_x") xnum_for(group2) + offset_for(fixed_level)
           else                               xnum_for(fixed_level) + offset_for(group2)
    ) %>%
    ungroup() %>%
    filter(x1 != x2) %>%   # degenerate (zero-width) brackets are not drawable
    arrange(pmin(x1, x2), abs(x2 - x1)) %>%
    mutate(y = base_y + (row_number() - 1) * step, x_mid = (x1 + x2) / 2)

  list(brackets = brackets, tip = tip)
}

# geom_sig_brackets(): ggplot layer list drawing a horizontal bracket + star
# for every row of a build_sig_brackets() result. Adds nothing when there are
# no significant comparisons, so it is always safe to add to a plot.
geom_sig_brackets <- function(bracket_result) {
  b <- bracket_result$brackets
  if (nrow(b) == 0) return(list())
  tip <- bracket_result$tip
  list(
    geom_segment(data = b, aes(x = x1, xend = x2, y = y, yend = y),
                 inherit.aes = FALSE, linewidth = 0.5),
    geom_segment(data = b, aes(x = x1, xend = x1, y = y, yend = y - tip),
                 inherit.aes = FALSE, linewidth = 0.5),
    geom_segment(data = b, aes(x = x2, xend = x2, y = y, yend = y - tip),
                 inherit.aes = FALSE, linewidth = 0.5),
    geom_text(data = b, aes(x = x_mid, y = y, label = sig),
              inherit.aes = FALSE, vjust = -0.3, size = 4.2, fontface = "bold")
  )
}

cat("Defined: sig_stars(), dodge_offset(), build_sig_brackets(), geom_sig_brackets()\n")
Defined: sig_stars(), dodge_offset(), build_sig_brackets(), geom_sig_brackets()

13.4 Within-family temperature comparisons, and within-temperature family comparisons

Four stratified pairwise comparisons, each via Wilcoxon rank-sum, with Holm correction applied across the family of four tests. This is written generically over the levels of family and temperature present in stats_data, so it extends unchanged to analyses with additional families or temperature levels: every same-family pair across temperature levels, and every same-temperature pair across families, is tested and Holm-corrected as one family of comparisons.

family_levels <- levels(stats_data$family)
temp_levels   <- levels(stats_data$temperature)

# Every pairwise comparison within a fixed family, across temperature levels.
# comparison_type = "across_x" -- generic tag consumed by build_sig_brackets()
# below: it means "x_var (temperature) varies, dodge_var (family) is fixed".
across_temp_tests <- bind_rows(lapply(family_levels, function(fam) {
  d <- filter(stats_data, family == fam)
  combn(temp_levels, 2, simplify = FALSE) %>%
    lapply(function(pair) {
      p <- wilcox.test(norm_glycogen ~ temperature,
                        data = filter(d, temperature %in% pair))$p.value
      data.frame(comparison_type = "across_x", fixed_level = fam,
                 group1 = pair[1], group2 = pair[2], p_value = p,
                 label_prefix = "within-family", stringsAsFactors = FALSE)
    }) %>% bind_rows()
}))

# Every pairwise comparison within a fixed temperature, across family levels.
# comparison_type = "across_dodge" -- "dodge_var (family) varies, x_var
# (temperature) is fixed".
across_family_tests <- bind_rows(lapply(temp_levels, function(tp) {
  d <- filter(stats_data, temperature == tp)
  combn(family_levels, 2, simplify = FALSE) %>%
    lapply(function(pair) {
      p <- wilcox.test(norm_glycogen ~ family,
                        data = filter(d, family %in% pair))$p.value
      data.frame(comparison_type = "across_dodge", fixed_level = tp,
                 group1 = pair[1], group2 = pair[2], p_value = p,
                 label_prefix = "within-temperature", stringsAsFactors = FALSE)
    }) %>% bind_rows()
}))

stratified_tests <- bind_rows(across_temp_tests, across_family_tests) %>%
  mutate(
    p_holm = p.adjust(p_value, method = "holm"),
    sig    = sig_stars(p_holm),
    label  = paste0(label_prefix, " (", fixed_level, "): ", group1, " vs. ", group2)
  )

cat("--- stratified_tests ---\n")
str(stratified_tests)
kable(stratified_tests %>% select(label, comparison_type, p_value, p_holm, sig), digits = 4)
cat("\nSignificant (Holm-corrected) stratified comparisons in this dataset:",
    sum(!is.na(stratified_tests$sig)), "of", nrow(stratified_tests), "\n")
--- stratified_tests ---
'data.frame':   4 obs. of  9 variables:
 $ comparison_type: chr  "across_x" "across_x" "across_dodge" "across_dodge"
 $ fixed_level    : chr  "Family 1" "Family 9" "ambient" "36C"
 $ group1         : chr  "ambient" "ambient" "Family 1" "Family 1"
 $ group2         : chr  "36C" "36C" "Family 9" "Family 9"
 $ p_value        : num  0.505 0.505 0.195 0.878
 $ label_prefix   : chr  "within-family" "within-family" "within-temperature" "within-temperature"
 $ p_holm         : num  1 1 0.779 1
 $ sig            : chr  NA NA NA NA
 $ label          : chr  "within-family (Family 1): ambient vs. 36C" "within-family (Family 9): ambient vs. 36C" "within-temperature (ambient): Family 1 vs. Family 9" "within-temperature (36C): Family 1 vs. Family 9"
label comparison_type p_value p_holm sig
within-family (Family 1): ambient vs. 36C across_x 0.5054 1.0000 NA
within-family (Family 9): ambient vs. 36C across_x 0.5054 1.0000 NA
within-temperature (ambient): Family 1 vs. Family 9 across_dodge 0.1949 0.7795 NA
within-temperature (36C): Family 1 vs. Family 9 across_dodge 0.8785 1.0000 NA 
Significant (Holm-corrected) stratified comparisons in this dataset: 0 of 4

14 EXPLORATORY DISTRIBUTION PLOT

The box plot is annotated with brackets for every statistically significant (Holm-corrected p < 0.05) pairwise comparison from stratified_tests above, via the reusable build_sig_brackets()/geom_sig_brackets() helpers. With the current dataset none of the four stratified comparisons are significant, so no brackets are drawn – the annotation layer is present but empty. If a future re-analysis (e.g. additional families, or a stronger temperature effect) produces a significant pairwise result, a bracket and asterisk(s) will appear automatically without any code changes.

box_brackets <- build_sig_brackets(stats_data, y_var = "norm_glycogen",
                                    x_var = "temperature", dodge_var = "family",
                                    comparisons = stratified_tests, dodge_width = 0.75)

cat("--- box_brackets$brackets (significant comparisons drawn on the box plot) ---\n")
str(box_brackets$brackets)

group_plot <- ggplot(stats_data, aes(x = temperature, y = norm_glycogen, fill = family)) +
  geom_boxplot(position = position_dodge(0.75), alpha = 0.35, outlier.shape = NA) +
  geom_point(aes(shape = family), position = position_jitterdodge(jitter.width = 0.15, dodge.width = 0.75),
             size = 2.8, alpha = 0.8) +
  geom_sig_brackets(box_brackets) +
  scale_y_continuous(expand = expansion(mult = c(0.05, 0.12))) +
  labs(x = "Temperature exposure", y = "Normalized glycogen (\u00b5g/mL/mg tissue)",
       title = "Glycogen content by family and temperature",
       subtitle = "Combined dataset (2026-07-30 + 2026-08-03 reconciled); boxplot + individual points",
       fill = "Family", shape = "Family") +
  scale_fill_manual(values = c("Family 1" = "#1b9e77", "Family 9" = "#d95f02")) +
  theme_bw(base_size = 13) +
  theme(axis.title = element_text(size = 11, face = "bold"),
        legend.position = "bottom")

print(group_plot)

ggsave(file.path(output_dir, "glycogen_by_family_temperature_combined.png"), group_plot,
       width = 8, height = 5.5, dpi = 300)
--- box_brackets$brackets (significant comparisons drawn on the box plot) ---
'data.frame':   0 obs. of  9 variables:
 $ comparison_type: chr 
 $ fixed_level    : chr 
 $ group1         : chr 
 $ group2         : chr 
 $ p_value        : num 
 $ label_prefix   : chr 
 $ p_holm         : num 
 $ sig            : chr 
 $ label          : chr

15 FAMILY x TEMPERATURE INTERACTION

The interaction plot is annotated using the same build_sig_brackets() helper, restricted to the "across_temperature" (within-family) comparisons and drawn with dodge_width = 0 since family is encoded by line color here rather than by horizontal dodging. Within-temperature family-vs-family comparisons are already shown as brackets on the box plot above rather than duplicated here, since family is not spatially separated at a given temperature in this plot.

interaction_data <- stats_data %>%
  group_by(family, temperature) %>%
  summarise(mean_glycogen = mean(norm_glycogen),
            se_glycogen   = sd(norm_glycogen) / sqrt(n()),
            .groups = "drop")

cat("--- interaction_data ---\n")
str(interaction_data)

interaction_brackets <- build_sig_brackets(
  interaction_data %>% mutate(norm_glycogen = mean_glycogen + se_glycogen),
  y_var = "norm_glycogen", x_var = "temperature", dodge_var = "family",
  comparisons = filter(stratified_tests, comparison_type == "across_x"),
  dodge_width = 0
)

cat("--- interaction_brackets$brackets (significant within-family comparisons drawn) ---\n")
str(interaction_brackets$brackets)

interaction_plot <- ggplot(interaction_data, aes(x = temperature, y = mean_glycogen,
                                                  color = family, group = family)) +
  geom_line(linewidth = 0.9) +
  geom_point(size = 3) +
  geom_errorbar(aes(ymin = mean_glycogen - se_glycogen, ymax = mean_glycogen + se_glycogen),
                width = 0.1) +
  geom_sig_brackets(interaction_brackets) +
  scale_y_continuous(expand = expansion(mult = c(0.05, 0.12))) +
  scale_color_manual(values = c("Family 1" = "#1b9e77", "Family 9" = "#d95f02")) +
  labs(x = "Temperature exposure", y = "Mean normalized glycogen (\u00b5g/mL/mg, \u00b1 SE)",
       title = "Family x temperature interaction",
       subtitle = "Combined dataset (2026-07-30 + 2026-08-03 reconciled); group means \u00b1 SE",
       color = "Family") +
  theme_bw(base_size = 13) +
  theme(axis.title = element_text(size = 11, face = "bold"),
        legend.position = "bottom")

print(interaction_plot)

ggsave(file.path(output_dir, "glycogen_family_temperature_interaction_combined.png"), interaction_plot,
       width = 7, height = 5.5, dpi = 300)
--- interaction_data ---
tibble [4 × 4] (S3: tbl_df/tbl/data.frame)
 $ family       : Factor w/ 2 levels "Family 1","Family 9": 1 1 2 2
 $ temperature  : Factor w/ 2 levels "ambient","36C": 1 2 1 2
 $ mean_glycogen: num [1:4] 30.2 21.1 16.5 22
 $ se_glycogen  : num [1:4] 8.26 5.12 4.48 4.82
--- interaction_brackets$brackets (significant within-family comparisons drawn) ---
'data.frame':   0 obs. of  9 variables:
 $ comparison_type: chr 
 $ fixed_level    : chr 
 $ group1         : chr 
 $ group2         : chr 
 $ p_value        : num 
 $ label_prefix   : chr 
 $ p_holm         : num 
 $ sig            : chr 
 $ label          : chr

16 SUMMARY

  • Combined dataset: 32 samples (families 1 and 9, ambient and 36°C), reconciling 2026-07-30 and 2026-08-03 measurements per the rule in Background. 5 samples had a measurement from both dates; for 4 of these (1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient) the 2026-08-03 (1:200) in-range value superseded the original extrapolated value, while 1_04_36C retained its 2026-07-30 (1:25) value since its 2026-08-03 re-assay fell below the standard curve’s floor.

  • Out-of-range samples: 0 sample(s) remain outside their standard curve’s quantifiable range after reconciliation and were excluded from all statistical comparisons above (see “Excluding out-of-range samples” for detail).

  • High-CV samples: samples exceeding 15% technical-replicate CV were flagged (see “Coefficient of Variation (CV) Check”) but retained in all analyses, per the analysis requirements.

  • Family effect: no statistically significant difference between Family 1 and Family 9 in normalized glycogen content, either pooled across temperature (two-way ANOVA family term, and the overall Wilcoxon test) or within either temperature stratum (Holm-corrected pairwise comparisons).

  • Temperature effect: no statistically significant difference between ambient and 36°C exposure, either pooled across family or within either family stratum.

  • Family x temperature interaction: no statistically significant interaction term in the two-way ANOVA. The interaction plot shows Family 1’s mean glycogen trending down from ambient to 36°C while Family 9’s trends up, but neither the interaction term nor the stratified within-family comparisons reach significance at this sample size – this pattern should be interpreted as a visual trend, not a confirmed effect.

]]> 2026 glycogen Glycogen-Glo SORMI Pacific oyster Magallana gigas Crassostrea gigas https://robertslab.github.io/sams-notebook/posts/2026/2026-08-06-Glycogen-Analysis---SORMI-June-2026-M.gigas-Family-1-vs-Family-9/ Thu, 06 Aug 2026 07:00:00 GMT Glycogen Assay - SORMI June 2026 M.gigas USDA Families 1 and 9 Sample Re-runs Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-08-03-Glycogen-Assay---SORMI-June-2026-M.gigas-USDA-Families-1-and-9-Sample-Re-runs/

INTRO

As part of the June 2026 SORMI project, I conducted a glycogen assay on Magallana gigas (Pacific oyster) ctenidia samples (ambient and 36°C temperatures) from USDA Families 1 and 9 (chosen by Steven). The purpose of this study was to compare the glycogen content between these two families using the Glycogen Glo Kit. Samples had been previously homogenized on 20260722 (Notebook entry) as well as assayed for glycogen on 20260730 (Notebook entry). However, four samples were outside of the standard curve and one had poor coefficient of variation (CV) between replicates, so a re-run of those samples was necessary.

  • 1_06_ambient

  • 1_08_ambient

  • 1_04_36C

  • 1_05_36C

  • 9_05_ambient

Important

Sample 1_04_36C was re-run due to poor CV between replicates in the previous assay attempt. It should have been processed with the previous 1:25 dilution, but was accidentally processed with a 1:200 dilution which resulted in a low signal.

All data/code/analysis is available here:

MATERIALS & METHODS

Homogenized ctenidia samples were thawed, vortexed, and spun at 16,000g for 1min to pellet debris. Based on the previous attempt at assaying these samples on 20260730 (Notebook) the following was decided upon:

  • dilution of 1:200 should be appropriate for samples to fall within standard curve range

  • glucose analysis is unnecessary for M.gigas ctenidia samples, as glucose has been shown to be negligible

Dilutions

All samples were diluted 1:200 in dilution buffer (2:1:1 ratio of PBS:0.3N HCl:TRIS). E.g. 1uL of sample + 199uL of dilution buffer. The diluted samples were then used for the glycogen assay.

Glycogen Assay

A standard curve of glycogen (20ug, 10ug, 2ug, 0.2ug, 0.02ug, 0ug) in PBS was prepared. The assay was performed in triplicate for each sample in a separate 96-well, white Costar plate, and luminescence readings were taken using a HTX Synergy plate reader (Agilent). A negative control (dilution buffer only), in triplicate, was included in the assay to account for background luminescence.

The assay was performed in triplicate for each sample in a 96-well, white Costar plate, and luminescence readings were taken using a HTX Synergy plate reader (Agilent).

The Glycogen Glo Kit protocol was followed according to the manufacturer’s instructions, with appropriate incubation times and reagent additions.

Master Mix Calcs

Digestion Master Mix

COMPONENT vol.per.rxn(uL) no.rxns total(uL)
glucoamylase buffer 25 36 900
glucoamylase 0.125 36 4.5
TOTAL 904.5

Detection Master Mix

COMPONENT vol.per.rxn(uL) no.rxns total(uL)
luciferin detection solution 50 36 1800
reductase substrate 0.25 36 9
reductase 0.25 36 9
NAD 0.25 36 9
glucose dehydrogenase 2 36 72
TOTAL 1899

Data Analysis

Data was analyzed in the following R Markdown file:

Analysis assessed the success of the new dilution factor (1:200).

RESULTS

A dilution factor of 1:200 was successful in bringing all samples within the standard curve range. Since I spaced out and also diluted the 1_04_36C sample (which should have been kept at the original 1:25 dilution), the signal was low and the sample was below the standard curve. However, the other four samples were all within the standard curve range and had good CVs between replicates. I should re-assay 1_04_36C again…

AI Use Level 4: Substantial AI contribution

Note

This section was knitted from Gen5-20260803-mgig-fams1_9-glycogen-glo.Rmd (GitHub).

1 BACKGROUND

Targeted re-assay of five Magallana gigas (Pacific oyster) ctenidia samples from the 2026-07-30 glycogen run (Gen5-20260730-mgig-fams1_9-glycogen-glo) using the Glycogen-Glo Assay (Promega) (GitHub; PDF), read on 2026-08-03.

Four of these five samples read above the top standard (20 µg/mL) in the original run and required extrapolation: 1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient. The fifth, 1_04_36C, was within range but had a technical-triplicate CV of 30.1%, above the 15% QC threshold used in the original analysis. All five were re-assayed here at a 1:200 dilution (vs. 1:25 originally, ~8× higher), using the same homogenate as the original run (tissue weights match exactly between runs).

This is a small, targeted follow-up analysis, not a full re-run of the 32-sample experiment. Results here are intended to supersede the corresponding five rows of the original run’s results table where the values differ meaningfully; see the Summary for a direct comparison and recommendation.

1.1 Sample naming

Per ../data/raw_luminescence/README.md, plate layout entries follow:

  • <sample>-<assay_type>-<tissue_weight>-df.<dilution_factor>

Here <sample> itself is composite:

  • <family>_<individual>_<temperature>

E.g. 1_06_ambient-glyc-11.0-df.200 is family 1, individual 06, ambient exposure, glycogen assay, 11.0 mg of ctenidia tissue, diluted 1:200.

Standards follow STD-<assay_type>-<concentration>, e.g. STD-glyc-20 (20 µg/mL glycogen); the buffer-only well is NEG-glyc.

1.2 Important note(s)

  1. Tissue weights are taken from the plate layout labels rather than a separate weights CSV, and match the corresponding weights in the original 2026-07-30 run exactly, confirming these are re-assays of the same homogenates rather than newly prepared samples.

2 SETUP

2.1 Libraries

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Output directory

# Output directory for this analysis (matches this file's name, per ../code/README.md)
output_dir <- "../output/Gen5-20260803-mgig-fams1_9-glycogen-glo"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- output_dir: destination for all figures and tables ---\n")
str(output_dir)
--- output_dir: destination for all figures and tables ---
 chr "../output/Gen5-20260803-mgig-fams1_9-glycogen-glo"

3 DATA

Data are read from the local repo (../data/raw_luminescence/) so this document renders before/after the files are pushed to GitHub. A single plate holds both the five re-assayed samples and the fresh standard curve/negative control.

data_dir <- "../data/raw_luminescence"

plate_layout <- read.csv(file.path(data_dir, "layout-Gen5-20260803-mgig-fams1_9-plate-01.csv"),
                          header = FALSE, stringsAsFactors = FALSE)
raw_luminescence <- read.csv(file.path(data_dir, "raw_lum-Gen5-20260803-mgig-fams1_9-plate-01.csv"),
                              header = FALSE)

cat("Plate layout:\n\n")
str(plate_layout)
cat("\nRaw luminescence:\n\n")
str(raw_luminescence)
Plate layout:

'data.frame':   8 obs. of  12 variables:
 $ V1 : chr  "9_05-ambient-glyc-9.2-df.200" "1_05_36C-glyc-25.7-df.200" "STD-glyc-20" "STD-glyc-0.02" ...
 $ V2 : chr  "9_05-ambient-glyc-9.2-df.200" "1_05_36C-glyc-25.7-df.200" "STD-glyc-20" "STD-glyc-0.02" ...
 $ V3 : chr  "9_05-ambient-glyc-9.2-df.200" "1_05_36C-glyc-25.7-df.200" "STD-glyc-20" "STD-glyc-0.02" ...
 $ V4 : chr  "1_06_ambient-glyc-11.0-df.200" "" "STD-glyc-10" "STD-glyc-0" ...
 $ V5 : chr  "1_06_ambient-glyc-11.0-df.200" "" "STD-glyc-10" "STD-glyc-0" ...
 $ V6 : chr  "1_06_ambient-glyc-11.0-df.200" "" "STD-glyc-10" "STD-glyc-0" ...
 $ V7 : chr  "1_08_ambient-glyc-24.8-df.200" "" "STD-glyc-2" "" ...
 $ V8 : chr  "1_08_ambient-glyc-24.8-df.200" "" "STD-glyc-2" "" ...
 $ V9 : chr  "1_08_ambient-glyc-24.8-df.200" "" "STD-glyc-2" "" ...
 $ V10: chr  "1_04_36C-glyc-12.4-df.200" "" "STD-glyc-0.2" "" ...
 $ V11: chr  "1_04_36C-glyc-12.4-df.200" "" "STD-glyc-0.2" "" ...
 $ V12: chr  "1_04_36C-glyc-12.4-df.200" "" "STD-glyc-0.2" "" ...

Raw luminescence:

'data.frame':   8 obs. of  12 variables:
 $ V1 : int  6528 8906 61432 804 570 7 8 3
 $ V2 : int  6498 8471 62800 631 576 5 8 6
 $ V3 : int  6599 8849 60190 608 604 7 4 5
 $ V4 : int  9137 29 28374 499 9 9 5 4
 $ V5 : int  9214 19 27652 418 4 5 3 4
 $ V6 : int  9304 16 28683 21 7 4 7 4
 $ V7 : int  31079 30 11790 10 7 4 7 5
 $ V8 : int  30545 27 12124 10 6 8 5 2
 $ V9 : int  30678 21 12465 11 9 5 4 2
 $ V10: int  833 12 3474 8 5 4 6 8
 $ V11: int  776 5 1771 6 6 4 3 4
 $ V12: int  860 6 1415 5 5 4 5 4

3.1 Reshape plate to long format

# Convert a paired layout/luminescence plate into one row per well
plate_to_long <- function(layout, luminescence, plate_label) {
  n_row <- 8
  n_col <- 12
  out <- expand.grid(plate_row_idx = 1:n_row, plate_col = 1:n_col)
  out$plate       <- plate_label
  out$plate_row   <- LETTERS[out$plate_row_idx]
  out$well        <- sprintf("%s%02d", out$plate_row, out$plate_col)
  out$label       <- trimws(as.character(mapply(function(i, j) layout[i, j],
                                                out$plate_row_idx, out$plate_col)))
  out$luminescence <- as.numeric(mapply(function(i, j) luminescence[i, j],
                                        out$plate_row_idx, out$plate_col))
  out <- out[order(out$plate_row_idx, out$plate_col), ]
  out[out$label != "" & !is.na(out$label), ]
}

plate1_long <- plate_to_long(plate_layout, raw_luminescence, "plate-01")

cat("--- plate_to_long(): layout + luminescence -> one row per well ---\n\n")
str(args(plate_to_long))

cat("\n--- plate1_long: one row per occupied well ---\n\n")
str(plate1_long)
--- plate_to_long(): layout + luminescence -> one row per well ---

function (layout, luminescence, plate_label)  

--- plate1_long: one row per occupied well ---

'data.frame':   36 obs. of  7 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "plate-01" "plate-01" "plate-01" "plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "1_06_ambient-glyc-11.0-df.200" ...
 $ luminescence : num  6528 6498 6599 9137 9214 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

4 STANDARD CURVE

4.1 Extract luminescence data

Row C holds the 20, 10, 2, and 0.2 µg/mL standards; row D holds the 0.02 and 0 µg/mL standards; row E columns 1-3 hold the NEG-glyc buffer-only negative control – all in triplicate, on the same plate as the re-assayed samples.

standards <- plate1_long %>%
  filter(grepl("^STD-glyc-", label)) %>%
  mutate(glyc_concentration = as.numeric(sub("^STD-glyc-", "", label)))

negative_control <- plate1_long %>% filter(grepl("^NEG-glyc", label))

cat("Standard concentrations (ug/mL):",
    paste(sort(unique(standards$glyc_concentration)), collapse = ", "), "\n")
cat("Replicates per standard:",
    paste(unique(table(standards$glyc_concentration)), collapse = ", "), "\n\n")

cat("Negative control (buffer only) luminescence:",
    paste(negative_control$luminescence, collapse = ", "),
    "| mean =", round(mean(negative_control$luminescence), 1), "\n")
cat("Zero glycogen standard luminescence:  mean =",
    round(mean(standards$luminescence[standards$glyc_concentration == 0]), 1), "\n")

cat("\n--- standards: standard curve wells with parsed concentration ---\n")
str(standards)
cat("\n--- negative_control: buffer-only wells ---\n")
str(negative_control)
Standard concentrations (ug/mL): 0, 0.02, 0.2, 2, 10, 20 
Replicates per standard: 3 

Negative control (buffer only) luminescence: 570, 576, 604 | mean = 583.3 
Zero glycogen standard luminescence:  mean = 312.7 

--- standards: standard curve wells with parsed concentration ---
'data.frame':   18 obs. of  8 variables:
 $ plate_row_idx     : int  3 3 3 3 3 3 3 3 3 3 ...
 $ plate_col         : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate             : chr  "plate-01" "plate-01" "plate-01" "plate-01" ...
 $ plate_row         : chr  "C" "C" "C" "C" ...
 $ well              : chr  "C01" "C02" "C03" "C04" ...
 $ label             : chr  "STD-glyc-20" "STD-glyc-20" "STD-glyc-20" "STD-glyc-10" ...
 $ luminescence      : num  61432 62800 60190 28374 27652 ...
 $ glyc_concentration: num  20 20 20 10 10 10 2 2 2 0.2 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- negative_control: buffer-only wells ---
'data.frame':   3 obs. of  7 variables:
 $ plate_row_idx: int  5 5 5
 $ plate_col    : int  1 2 3
 $ plate        : chr  "plate-01" "plate-01" "plate-01"
 $ plate_row    : chr  "E" "E" "E"
 $ well         : chr  "E01" "E02" "E03"
 $ label        : chr  "NEG-glyc" "NEG-glyc" "NEG-glyc"
 $ luminescence : num  570 576 604
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

The buffer-only negative control and the zero-glycogen standard are again in close agreement (913 vs. 910 RLU in the original run; see below), as expected.

4.2 Glycogen standard curve summary statistics and linear regression

glycogen_summary_data <- standards %>%
  group_by(glyc_concentration) %>%
  summarise(
    glyc_mean_luminescence = mean(luminescence),
    glyc_sd                = sd(luminescence),
    glyc_se                = sd(luminescence) / sqrt(n()),
    glyc_cv                = 100 * sd(luminescence) / mean(luminescence),
    glyc_n                 = n(),
    .groups = "drop"
  ) %>%
  arrange(glyc_concentration)

lm_model      <- lm(glyc_mean_luminescence ~ glyc_concentration, data = glycogen_summary_data)
glyc_slope     <- coef(lm_model)[2]
glyc_intercept <- coef(lm_model)[1]
glyc_r_squared <- summary(lm_model)$r.squared

lm_model_reps    <- lm(luminescence ~ glyc_concentration, data = standards)
glyc_r2_reps     <- summary(lm_model_reps)$r.squared

glyc_conc_min_nonzero <- min(glycogen_summary_data$glyc_concentration[
  glycogen_summary_data$glyc_concentration > 0])
glyc_conc_max <- max(glycogen_summary_data$glyc_concentration)

kable(glycogen_summary_data,
      digits = c(2, 1, 1, 1, 2, 0),
      col.names = c("Glycogen (ug/mL)", "Mean luminescence", "SD", "SEM", "CV (%)", "n"),
      caption = "Glycogen standard curve summary statistics (this plate)")

cat("\nFit to concentration means:      y =", sprintf("%.1f", glyc_slope), "x +",
    sprintf("%.1f", glyc_intercept), " R^2 =", sprintf("%.4f", glyc_r_squared), "\n")
cat("Fit to individual replicates:    R^2 =", sprintf("%.4f", glyc_r2_reps), "\n")
cat("Quantifiable range (standards):", glyc_conc_min_nonzero, "-", glyc_conc_max, "ug/mL\n")

cat("\n--- glycogen_summary_data: per-concentration summary statistics ---\n")
str(glycogen_summary_data)
cat("\n--- lm_model: regression on per-concentration means ---\n")
str(lm_model, max.level = 1, give.attr = FALSE)
Glycogen (ug/mL) Mean luminescence SD SEM CV (%) n
0.00 312.7 255.8 147.7 81.82 3
0.02 681.0 107.1 61.9 15.73 3
0.20 2220.0 1100.5 635.4 49.57 3
2.00 12126.3 337.5 194.9 2.78 3
10.00 28236.3 529.1 305.5 1.87 3
20.00 61474.0 1305.5 753.7 2.12 3

Glycogen standard curve summary statistics (this plate)

Fit to concentration means:      y = 2937.7 x + 1733.0  R^2 = 0.9886 
Fit to individual replicates:    R^2 = 0.9878 
Quantifiable range (standards): 0.02 - 20 ug/mL

--- glycogen_summary_data: per-concentration summary statistics ---
tibble [6 × 6] (S3: tbl_df/tbl/data.frame)
 $ glyc_concentration    : num [1:6] 0 0.02 0.2 2 10 20
 $ glyc_mean_luminescence: num [1:6] 313 681 2220 12126 28236 ...
 $ glyc_sd               : num [1:6] 256 107 1100 338 529 ...
 $ glyc_se               : num [1:6] 147.7 61.9 635.4 194.9 305.5 ...
 $ glyc_cv               : num [1:6] 81.82 15.73 49.57 2.78 1.87 ...
 $ glyc_n                : int [1:6] 3 3 3 3 3 3

--- lm_model: regression on per-concentration means ---
List of 12
 $ coefficients : Named num [1:2] 1733 2938
 $ residuals    : Named num [1:6] -1420 -1111 -100 4518 -2874 ...
 $ effects      : Named num [1:6] -42887 53448 426 4987 -2659 ...
 $ rank         : int 2
 $ fitted.values: Named num [1:6] 1733 1792 2320 7608 31110 ...
 $ assign       : int [1:2] 0 1
 $ qr           :List of 5
 $ df.residual  : int 4
 $ xlevels      : Named list()
 $ call         : language lm(formula = glyc_mean_luminescence ~ glyc_concentration, data = glycogen_summary_data)
 $ terms        :Classes 'terms', 'formula'  language glyc_mean_luminescence ~ glyc_concentration
 $ model        :'data.frame':  6 obs. of  2 variables:

Standard curve QC note. The 0 and 0.2 µg/mL standards show high triplicate CV (81.8% and 49.6% respectively) – both are low-signal wells where a small absolute RLU difference translates to a large percentage swing, and one 0 µg/mL well (21 RLU) reads well below its two replicates (499, 418 RLU). Excluding that single well changes the fitted slope/intercept by <2%, and has a negligible effect on back-calculated concentrations for the samples below (e.g. 1_04_36C’s well concentration shifts from -0.310 to -0.323 µg/mL) – this variability at the assay’s low end does not materially affect the regression, but it is consistent with 1_04_36C reading close enough to background that its sign is not reliable (see Out-of-range / floor-effect check below). The curve overall fits tightly (R^2 = 0.9886, individual-replicate R^2 = 0.9878), consistent with the original run’s standard curve quality.

4.3 Extract sample data and calculate glycogen levels

samples_wells <- plate1_long %>%
  filter(!grepl("^STD-glyc-", label) & !grepl("^NEG-glyc", label)) %>%
  # Normalize `-` to `_` in the <family>-<individual>-<temperature> prefix
  mutate(label_clean = sub("^([19])[-_]([0-9]{2})[-_](ambient|36C)-", "\\1_\\2_\\3-", label)) %>%
  mutate(
    sample_id  = sub("-glyc-.*$", "", label_clean),
    family     = sub("^([19])_.*$", "\\1", label_clean),
    individual = sub("^[19]_([0-9]{2})_.*$", "\\1", label_clean),
    temperature = sub("^[19]_[0-9]{2}_([^-]+)-.*$", "\\1", label_clean),
    weight_mg  = as.numeric(sub("^.*-glyc-([0-9.]+)-df\\..*$", "\\1", label_clean)),
    dilution   = as.numeric(sub("^.*-df\\.", "", label_clean))
  ) %>%
  mutate(
    family      = factor(paste("Family", family), levels = c("Family 1", "Family 9")),
    temperature = factor(temperature, levels = c("ambient", "36C"))
  )

# Fail loudly rather than silently dropping malformed labels
stopifnot(!any(is.na(samples_wells$weight_mg)),
          !any(is.na(samples_wells$dilution)),
          !any(is.na(samples_wells$temperature)))

cat("Wells parsed:", nrow(samples_wells),
    "| unique samples:", length(unique(samples_wells$sample_id)), "\n")
cat("Dilution factor(s) used:", paste(unique(samples_wells$dilution), collapse = ", "), "\n\n")
cat("Samples re-assayed:", paste(sort(unique(samples_wells$sample_id)), collapse = ", "), "\n")
cat("Tissue weight range (mg):",
    paste(range(samples_wells$weight_mg), collapse = " - "), "\n")

cat("\n--- samples_wells: one row per sample well, metadata parsed from label ---\n")
str(samples_wells)
Wells parsed: 15 | unique samples: 5 
Dilution factor(s) used: 200 

Samples re-assayed: 1_04_36C, 1_05_36C, 1_06_ambient, 1_08_ambient, 9_05_ambient 
Tissue weight range (mg): 9.2 - 25.7 

--- samples_wells: one row per sample well, metadata parsed from label ---
'data.frame':   15 obs. of  14 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "plate-01" "plate-01" "plate-01" "plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "9_05-ambient-glyc-9.2-df.200" "1_06_ambient-glyc-11.0-df.200" ...
 $ luminescence : num  6528 6498 6599 9137 9214 ...
 $ label_clean  : chr  "9_05_ambient-glyc-9.2-df.200" "9_05_ambient-glyc-9.2-df.200" "9_05_ambient-glyc-9.2-df.200" "1_06_ambient-glyc-11.0-df.200" ...
 $ sample_id    : chr  "9_05_ambient" "9_05_ambient" "9_05_ambient" "1_06_ambient" ...
 $ family       : Factor w/ 2 levels "Family 1","Family 9": 2 2 2 1 1 1 1 1 1 1 ...
 $ individual   : chr  "05" "05" "05" "06" ...
 $ temperature  : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 1 2 ...
 $ weight_mg    : num  9.2 9.2 9.2 11 11 11 24.8 24.8 24.8 12.4 ...
 $ dilution     : num  200 200 200 200 200 200 200 200 200 200 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

Per-sample values are computed identically to the original analysis: mean of the technical triplicate, back-calculated through this plate’s own standard curve.

  1. Well concentration = (mean luminescence - intercept) / slope → µg/mL in the assayed (diluted) well.
  2. Homogenate concentration = well concentration × dilution factor (200) → µg/mL in the undiluted homogenate.
  3. Normalized glycogen = homogenate concentration / tissue weight → µg/mL per mg tissue.
sample_glycogen <- samples_wells %>%
  group_by(sample_id, family, individual, temperature, weight_mg, dilution) %>%
  summarise(
    n_reps        = n(),
    mean_lum      = mean(luminescence),
    sd_lum        = sd(luminescence),
    cv_lum        = 100 * sd(luminescence) / mean(luminescence),
    .groups = "drop"
  ) %>%
  mutate(
    well_conc_ug_mL       = (mean_lum - glyc_intercept) / glyc_slope,
    homogenate_conc_ug_mL = well_conc_ug_mL * dilution,
    norm_glycogen         = homogenate_conc_ug_mL / weight_mg,
    in_std_range          = well_conc_ug_mL >= glyc_conc_min_nonzero &
                            well_conc_ug_mL <= glyc_conc_max
  ) %>%
  arrange(family, temperature, individual)

cat("Samples quantified:", nrow(sample_glycogen), "\n")

cat("\n--- sample_glycogen: one row per individual, triplicate averaged and back-calculated ---\n")
str(sample_glycogen)
Samples quantified: 5 

--- sample_glycogen: one row per individual, triplicate averaged and back-calculated ---
tibble [5 × 14] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:5] "1_06_ambient" "1_08_ambient" "1_04_36C" "1_05_36C" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 2
 $ individual           : chr [1:5] "06" "08" "04" "05" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 2 2 1
 $ weight_mg            : num [1:5] 11 24.8 12.4 25.7 9.2
 $ dilution             : num [1:5] 200 200 200 200 200
 $ n_reps               : int [1:5] 3 3 3 3 3
 $ mean_lum             : num [1:5] 9218 30767 823 8742 6542
 $ sd_lum               : num [1:5] 83.6 278 42.9 236.4 51.9
 $ cv_lum               : num [1:5] 0.907 0.903 5.211 2.704 0.793
 $ well_conc_ug_mL      : num [1:5] 2.55 9.88 -0.31 2.39 1.64
 $ homogenate_conc_ug_mL: num [1:5] 510 1977 -62 477 327
 $ norm_glycogen        : num [1:5] 46.3 79.7 -5 18.6 35.6
 $ in_std_range         : logi [1:5] TRUE TRUE FALSE TRUE TRUE

4.4 Plot standard curve with re-assayed sample points

std_curve_plot <- ggplot(glycogen_summary_data,
                         aes(x = glyc_concentration, y = glyc_mean_luminescence)) +
  geom_smooth(method = "lm", se = FALSE, color = "grey60", linetype = "dashed") +
  geom_errorbar(aes(ymin = glyc_mean_luminescence - glyc_se,
                    ymax = glyc_mean_luminescence + glyc_se), width = 0.3) +
  geom_point(size = 3, color = "#3182bd") +
  geom_point(data = sample_glycogen,
             aes(x = well_conc_ug_mL, y = mean_lum, shape = in_std_range),
             color = "#d94801", size = 3, inherit.aes = FALSE) +
  scale_shape_manual(values = c(`TRUE` = 16, `FALSE` = 4),
                     labels = c(`TRUE` = "in range", `FALSE` = "out of range / floor"),
                     name = "Re-assayed sample") +
  labs(title = "Glycogen standard curve (this plate) with re-assayed sample points",
       x = "Glycogen (ug/mL)", y = "Mean luminescence (RLU)") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"))

cat("--- std_curve_plot: ggplot object structure ---\n\n")
summary(std_curve_plot)

ggsave(file.path(output_dir, "glycogen_standard_curve_rerun.png"), std_curve_plot,
       width = 10, height = 6, dpi = 300)

std_curve_plot

--- std_curve_plot: ggplot object structure ---

data: glyc_concentration, glyc_mean_luminescence, glyc_sd, glyc_se,
  glyc_cv, glyc_n [6x6]
mapping:  x = ~glyc_concentration, y = ~glyc_mean_luminescence
scales:   shape 
faceting:  <empty> 
-----------------------------------
geom_smooth: na.rm = FALSE, orientation = NA, se = FALSE
stat_smooth: na.rm = FALSE, orientation = NA, se = FALSE, method = lm
position_identity 

mapping: ymin = ~glyc_mean_luminescence - glyc_se, ymax = ~glyc_mean_luminescence + glyc_se 
geom_errorbar: na.rm = FALSE, orientation = NA, lineend = butt, width = 0.3
stat_identity: na.rm = FALSE
position_identity 

geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: x = ~well_conc_ug_mL, y = ~mean_lum, shape = ~in_std_range 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

5 QUALITY CONTROL

5.1 Technical replicate variability

cat("Coefficient of variation across technical triplicates (%):\n")
print(round(sample_glycogen$cv_lum, 2))

cat("\nSamples with CV > 15%:\n")
high_cv <- sample_glycogen %>% filter(cv_lum > 15) %>%
  select(sample_id, mean_lum, sd_lum, cv_lum)
if (nrow(high_cv) == 0) {
  cat("  none\n")
} else {
  print(as.data.frame(high_cv), row.names = FALSE, digits = 4)
}

cat("\n--- high_cv: samples exceeding the 15% triplicate CV threshold ---\n\n")
str(high_cv)
Coefficient of variation across technical triplicates (%):
[1] 0.91 0.90 5.21 2.70 0.79

Samples with CV > 15%:
  none

--- high_cv: samples exceeding the 15% triplicate CV threshold ---

tibble [0 × 4] (S3: tbl_df/tbl/data.frame)
 $ sample_id: chr(0) 
 $ mean_lum : num(0) 
 $ sd_lum   : num(0) 
 $ cv_lum   : num(0)

1_04_36C – re-assayed specifically because of a 30.1% CV in the original run – now has a CV of 5.2%, well below threshold: the higher dilution resolved the reproducibility problem it was re-run for. See below for why this improvement in precision did not translate into a usable concentration estimate.

5.2 Out-of-range / floor-effect check

oor <- sample_glycogen %>%
  select(sample_id, family, temperature, mean_lum, well_conc_ug_mL,
         weight_mg, norm_glycogen, in_std_range) %>%
  arrange(desc(well_conc_ug_mL))

kable(oor, digits = c(0, 0, 0, 0, 3, 1, 2, 0),
      col.names = c("Sample", "Family", "Temp", "Mean lum.", "Well conc. (ug/mL)",
                    "Tissue (mg)", "Normalized glycogen", "In std range"),
      caption = "Re-assayed samples: back-calculated concentrations at 1:200 dilution")

cat("\n--- Full re-assay results, all 5 samples ---\n\n")
str(oor)
Sample Family Temp Mean lum. Well conc. (ug/mL) Tissue (mg) Normalized glycogen In std range
1_08_ambient Family 1 ambient 30767 9.883 24.8 79.70 TRUE
1_06_ambient Family 1 ambient 9218 2.548 11.0 46.33 TRUE
1_05_36C Family 1 36C 8742 2.386 25.7 18.57 TRUE
9_05_ambient Family 9 ambient 6542 1.637 9.2 35.58 TRUE
1_04_36C Family 1 36C 823 -0.310 12.4 -5.00 FALSE

Re-assayed samples: back-calculated concentrations at 1:200 dilution

--- Full re-assay results, all 5 samples ---

tibble [5 × 8] (S3: tbl_df/tbl/data.frame)
 $ sample_id      : chr [1:5] "1_08_ambient" "1_06_ambient" "1_05_36C" "9_05_ambient" ...
 $ family         : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 2 1
 $ temperature    : Factor w/ 2 levels "ambient","36C": 1 1 2 1 2
 $ mean_lum       : num [1:5] 30767 9218 8742 6542 823
 $ well_conc_ug_mL: num [1:5] 9.88 2.55 2.39 1.64 -0.31
 $ weight_mg      : num [1:5] 24.8 11 25.7 9.2 12.4
 $ norm_glycogen  : num [1:5] 79.7 46.3 18.6 35.6 -5
 $ in_std_range   : logi [1:5] TRUE TRUE TRUE TRUE FALSE

Four of the five samples (1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient) now fall comfortably within the standard curve’s quantifiable range (0.02-20 µg/mL) at 1:200 – the 8× dilution increase successfully resolved the original extrapolation problem for all four.

The fifth, 1_04_36C, was not originally out-of-range (it was re-assayed for high CV, not for reading above the curve); at 1:200 its mean luminescence (823 RLU) falls below the fitted intercept, producing a negative back-calculated well concentration. It was accidentally diluted at 1:200 for this re-assay, when it should have been re-assayed with its original dilution of 1:25.

6 COMPARISON TO ORIGINAL RUN

original_path <- "../output/Gen5-20260730-mgig-fams1_9-glycogen-glo/sample_glycogen_results.csv"
original <- read.csv(original_path)

original_sub <- original %>%
  filter(Sample %in% sample_glycogen$sample_id) %>%
  transmute(
    sample_id             = Sample,
    dilution_orig         = Dilution,
    mean_lum_orig         = Mean.lum.,
    cv_orig               = CV....,
    norm_glycogen_orig    = Normalized.glycogen..ug.mL.mg.,
    in_range_orig         = In.std.range
  )

rerun_sub <- sample_glycogen %>%
  transmute(
    sample_id          = sample_id,
    dilution_rerun     = dilution,
    mean_lum_rerun     = round(mean_lum, 0),
    cv_rerun           = round(cv_lum, 1),
    norm_glycogen_rerun = round(norm_glycogen, 2),
    in_range_rerun     = ifelse(in_std_range, "yes", "NO - extrapolated")
  )

comparison <- inner_join(original_sub, rerun_sub, by = "sample_id") %>%
  arrange(sample_id)

kable(comparison,
      col.names = c("Sample", "Dilution (orig)", "Mean lum. (orig)", "CV % (orig)",
                    "Norm. glycogen (orig)", "In range (orig)",
                    "Dilution (re-run)", "Mean lum. (re-run)", "CV % (re-run)",
                    "Norm. glycogen (re-run)", "In range (re-run)"),
      caption = "Original run (2026-07-30, 1:25) vs. targeted re-assay (2026-08-03, 1:200)")

cat("\n--- comparison: original vs. re-assay results, side by side ---\n\n")
str(comparison)
Sample Dilution (orig) Mean lum. (orig) CV % (orig) Norm. glycogen (orig) In range (orig) Dilution (re-run) Mean lum. (re-run) CV % (re-run) Norm. glycogen (re-run) In range (re-run)
1_04_36C 25 5075 30.1 2.83 yes 200 823 5.2 -5.00 NO - extrapolated
1_05_36C 25 94104 4.9 29.86 NO - extrapolated 200 8742 2.7 18.57 yes
1_06_ambient 25 97291 6.0 72.15 NO - extrapolated 200 9218 0.9 46.33 yes
1_08_ambient 25 200961 0.2 66.38 NO - extrapolated 200 30767 0.9 79.70 yes
9_05_ambient 25 70209 1.1 62.05 NO - extrapolated 200 6542 0.8 35.58 yes

Original run (2026-07-30, 1:25) vs. targeted re-assay (2026-08-03, 1:200)

--- comparison: original vs. re-assay results, side by side ---

'data.frame':   5 obs. of  11 variables:
 $ sample_id          : chr  "1_04_36C" "1_05_36C" "1_06_ambient" "1_08_ambient" ...
 $ dilution_orig      : int  25 25 25 25 25
 $ mean_lum_orig      : int  5075 94104 97291 200961 70209
 $ cv_orig            : num  30.1 4.9 6 0.2 1.1
 $ norm_glycogen_orig : num  2.83 29.86 72.15 66.38 62.05
 $ in_range_orig      : chr  "yes" "NO - extrapolated" "NO - extrapolated" "NO - extrapolated" ...
 $ dilution_rerun     : num  200 200 200 200 200
 $ mean_lum_rerun     : num  823 8742 9218 30767 6542
 $ cv_rerun           : num  5.2 2.7 0.9 0.9 0.8
 $ norm_glycogen_rerun: num  -5 18.6 46.3 79.7 35.6
 $ in_range_rerun     : chr  "NO - extrapolated" "yes" "yes" "yes" ...

6.1 Plot: original vs. re-assayed normalized glycogen

comparison_long <- comparison %>%
  select(sample_id, in_range_orig, in_range_rerun,
         norm_glycogen_orig, norm_glycogen_rerun) %>%
  pivot_longer(c(norm_glycogen_orig, norm_glycogen_rerun),
               names_to = "run", values_to = "norm_glycogen") %>%
  mutate(
    run = recode(run, norm_glycogen_orig = "Original (1:25)", norm_glycogen_rerun = "Re-assay (1:200)"),
    run = factor(run, levels = c("Original (1:25)", "Re-assay (1:200)")),
    in_range = ifelse(run == "Original (1:25)", in_range_orig, in_range_rerun),
    in_range = ifelse(in_range == "yes", "in range", "out of range / floor")
  )

comparison_plot <- ggplot(comparison_long, aes(x = run, y = norm_glycogen, group = sample_id)) +
  geom_line(color = "grey60", linewidth = 0.6) +
  geom_point(aes(color = sample_id, shape = in_range), size = 3.2) +
  scale_shape_manual(values = c("in range" = 16, "out of range / floor" = 4)) +
  labs(title = "Normalized glycogen: original run vs. targeted re-assay",
       subtitle = "Points connected by sample; 'x' marks values flagged out-of-range or below the detection floor",
       x = NULL, y = "Normalized glycogen (ug/mL/mg tissue)", color = "Sample", shape = "Status") +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        legend.position = "right")

cat("--- comparison_plot: ggplot object structure ---\n\n")
summary(comparison_plot)

ggsave(file.path(output_dir, "original_vs_rerun_comparison.png"), comparison_plot,
       width = 9, height = 5.5, dpi = 300)

comparison_plot

--- comparison_plot: ggplot object structure ---

data: sample_id, in_range_orig, in_range_rerun, run, norm_glycogen,
  in_range [10x6]
mapping:  x = ~run, y = ~norm_glycogen, group = ~sample_id
scales:   shape 
faceting:  <empty> 
-----------------------------------
geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: colour = ~sample_id, shape = ~in_range 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

7 SAMPLE GLYCOGEN TABLE

sample_table <- sample_glycogen %>%
  transmute(
    Sample        = sample_id,
    Family        = family,
    Temperature   = temperature,
    Individual    = individual,
    `Tissue (mg)` = weight_mg,
    `Dilution`    = dilution,
    `Mean lum.`   = round(mean_lum, 0),
    `CV (%)`      = round(cv_lum, 1),
    `Well glycogen (ug/mL)`        = round(well_conc_ug_mL, 3),
    `Homogenate glycogen (ug/mL)`  = round(homogenate_conc_ug_mL, 2),
    `Normalized glycogen (ug/mL/mg)` = round(norm_glycogen, 2),
    `In std range` = ifelse(in_std_range, "yes", "NO - extrapolated")
  )

kable(sample_table, caption = "Per-sample glycogen quantification, targeted re-assay (mean of technical triplicate)")

write.csv(sample_table, file.path(output_dir, "sample_glycogen_results.csv"), row.names = FALSE)

cat("\n--- sample_table: formatted per-sample results written to CSV ---\n\n")
str(sample_table)
Sample Family Temperature Individual Tissue (mg) Dilution Mean lum. CV (%) Well glycogen (ug/mL) Homogenate glycogen (ug/mL) Normalized glycogen (ug/mL/mg) In std range
1_06_ambient Family 1 ambient 06 11.0 200 9218 0.9 2.548 509.61 46.33 yes
1_08_ambient Family 1 ambient 08 24.8 200 30767 0.9 9.883 1976.68 79.70 yes
1_04_36C Family 1 36C 04 12.4 200 823 5.2 -0.310 -61.95 -5.00 NO - extrapolated
1_05_36C Family 1 36C 05 25.7 200 8742 2.7 2.386 477.18 18.57 yes
9_05_ambient Family 9 ambient 05 9.2 200 6542 0.8 1.637 327.38 35.58 yes

Per-sample glycogen quantification, targeted re-assay (mean of technical triplicate)

--- sample_table: formatted per-sample results written to CSV ---

tibble [5 × 12] (S3: tbl_df/tbl/data.frame)
 $ Sample                        : chr [1:5] "1_06_ambient" "1_08_ambient" "1_04_36C" "1_05_36C" ...
 $ Family                        : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 2
 $ Temperature                   : Factor w/ 2 levels "ambient","36C": 1 1 2 2 1
 $ Individual                    : chr [1:5] "06" "08" "04" "05" ...
 $ Tissue (mg)                   : num [1:5] 11 24.8 12.4 25.7 9.2
 $ Dilution                      : num [1:5] 200 200 200 200 200
 $ Mean lum.                     : num [1:5] 9218 30767 823 8742 6542
 $ CV (%)                        : num [1:5] 0.9 0.9 5.2 2.7 0.8
 $ Well glycogen (ug/mL)         : num [1:5] 2.55 9.88 -0.31 2.39 1.64
 $ Homogenate glycogen (ug/mL)   : num [1:5] 510 1977 -62 477 327
 $ Normalized glycogen (ug/mL/mg): num [1:5] 46.3 79.7 -5 18.6 35.6
 $ In std range                  : chr [1:5] "yes" "yes" "NO - extrapolated" "yes" ...

8 SUMMARY

What this re-assay was for. Five samples from the 2026-07-30 run were re-assayed at a higher dilution (1:200 vs. 1:25): four that read above the standard curve’s top standard (1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient) and one with a high technical-replicate CV (1_04_36C, 30.1% originally).

Standard curve. This plate’s standard curve fit well (y = 2937.7 x + 1733.0, R^2 = 0.9886), comparable in quality to the original run’s curve.

Dilution outcome: 4 of 5 successful. The 8× dilution increase worked as intended for all four samples that were originally out-of-range – all four now fall within the standard curve’s quantifiable range, with comfortable margin below the top standard (well concentrations 1.6-9.9 µg/mL vs. the 0.02-20 µg/mL curve). Their re-assayed normalized glycogen values are lower than the original (extrapolated) estimates in every case, consistent with extrapolation beyond a standard curve’s top point tending to overestimate concentration:

Sample Original (1:25, extrapolated) Re-assay (1:200, in range)
1_08_ambient 66.4 79.7
1_06_ambient 72.2 46.3
1_05_36C 29.9 18.6
9_05_ambient 62.1 35.6

(Note 1_08_ambient is the exception – its re-assayed value is higher than the original extrapolated estimate, underscoring that extrapolation error is not predictably one-directional and is exactly why re-assay within range matters rather than trusting the extrapolated numbers.)

Sample 1_04_36C should NOT have been diluted, as it previously failed QC due to a high CV. This should be re-assayed.

Using these results going forward. For 1_08_ambient, 1_06_ambient, 1_05_36C, and 9_05_ambient, the values in this document’s sample_glycogen_results.csv should replace the corresponding rows from the 2026-07-30 run in any downstream analysis (e.g. the family/temperature ANOVA), since they are now interpolated rather than extrapolated and were measured with a same-plate standard curve.

]]> 2026 glycogen Glycogen-Glo SORMI Pacific oyster Magallana gigas Crassostrea gigas https://robertslab.github.io/sams-notebook/posts/2026/2026-08-03-Glycogen-Assay---SORMI-June-2026-M.gigas-USDA-Families-1-and-9-Sample-Re-runs/ Mon, 03 Aug 2026 07:00:00 GMT Glycogen Assay - SORMI June 2026 M.gigas USDA Families 1 and 9 Comparisons Using Glycogen Glo Kit Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-07-30-Glycogen-Assay---SORMI-June-2026-M.gigas-USDA-Families-1-and-9-Comparisons-Using-Glycogen-Glo-Kit/

INTRO

As part of the June 2026 SORMI project, I conducted a glycogen assay on Magallana gigas (Pacific oyster) ctenidia samples (ambient and 36°C temperatures) from USDA Families 1 and 9 (chosen by Steven). The purpose of this study was to compare the glycogen content between these two families using the Glycogen Glo Kit. Samples had been previously homogenized on 20260722 (Notebook entry).

All data/code/analysis is available here:

MATERIALS & METHODS

Homogenized ctenidia samples were thawed, vortexed, and spun at 16,000g for 1min to pellet debris. Based on prior work by Hazel (Notebook) the following was decided upon:

  • dilution of 1:25 should be appropriate for samples to fall within standard curve range

  • glucose analysis is unnecessary for M.gigas ctenidia samples, as glucose has been shown to be negligible

Dilutions

All samples were diluted 1:25 in dilution buffer (2:1:1 ratio of PBS:0.3N HCl:TRIS). E.g. 4uL of sample + 96uL of dilution buffer. The diluted samples were then used for the glycogen assay.

Glycogen Assay

A standard curve of glycogen (20ug, 2ug, 0.2ug, 0.02ug, 0ug) in PBS was prepared. The assay was performed in triplicate for each sample in a separate 96-well, white Costar plate, and luminescence readings were taken using a HTX Synergy plate reader (Agilent). A negative control (dilution buffer only), in triplicate, was included in the assay to account for background luminescence.

The assay was performed in triplicate for each sample in a 96-well, white Costar plates, and luminescence readings were taken using a HTX Synergy plate reader (Agilent).

The Glycogen Glo Kit protocol was followed according to the manufacturer’s instructions, with appropriate incubation times and reagent additions.

Master Mix Calcs

Digestion Master Mix

COMPONENT vol.per.rxn(uL) no.rxns total(uL)
glucoamylase buffer 25 125 3125
glucoamylase 0.125 125 15.625
TOTAL 3140.625

Detection Master Mix

COMPONENT vol.per.rxn(uL) no.rxns total(uL)
luciferin detection solution 50 125 6250
reductase substrate 0.25 125 31.25
reductase 0.25 125 31.25
NAD 0.25 125 31.25
glucose dehydrogenase 2 125 250
TOTAL 6593.75

Data Analysis

Data was analyzed in the following R Markdown file:

See the BACKGROUND section below for code/details on the data analysis workflow.

RESULTS

None of the family, temperature, or family × temperature effects are statistically significant at n = 8 per group – excluding the four extrapolated samples changes this picture only slightly (log-scale ANOVA family p = 0.652, temperature p = 0.467, interaction p = 0.292, n = 28), so the qualitative conclusion (no detectable family or temperature effect) is robust to that exclusion.

I need to re-run the assay for the four extrapolated samples with higher dilutions to ensure they are within the standard curve range and then re-analyze the data. The current analysis is based on extrapolated values for those four samples, which may not be reliable.

AI Use Level 4: Substantial AI contribution

Note

This section was knitted from Gen5-20260730-mgig-fams1_9-glycogen-glo.Rmd (GitHub)

1 BACKGROUND

Glycogen quantification of Magallana gigas (Pacific oyster) ctenidia homogenates using the Glycogen-Glo Assay (Promega) (GitHub; PDF), read on 2026-07-30.

The experiment compares two USDA oyster families (family 1 and family 9) across two temperature exposures (ambient and 36°C), with 8 individuals per family per temperature (32 samples total). Each sample was run in technical triplicate at a 1:25 dilution.

Two plates were read on the same instrument/protocol:

  • Plate 01: all 32 samples, triplicate, filling all 96 wells (no standards).
  • Plate 02: glycogen standard curve (20, 2, 0.2, 0.02, 0 µg/mL) plus a buffer-only negative control, all in triplicate.

Because the standards live on a separate plate from the samples, a single standard curve from plate 02 is applied to all samples on plate 01. Both plates were read with the same protocol (luminesence-96wells-1.0s-glycogen_glo.prt, gain 135, 1.00 s integration, extended dynamic range) back-to-back, so this is reasonable, but it does mean plate-to-plate signal drift is not controlled for. This is flagged in the interpretation.

1.1 Sample naming

Per ../data/raw_luminescence/README.md, plate layout entries follow:

  • <sample>-<assay_type>-<tissue_weight>-df.<dilution_factor>

Here <sample> itself is composite:

  • <family>_<individual>_<temperature>

E.g. 1_05_36C-glyc-25.7-df.25 is family 1, individual 05, 36°C exposure, glycogen assay, 25.7 mg of ctenidia tissue, diluted 1:25.

Standards follow STD-<assay_type>-<concentration>, e.g. STD-glyc-20 (20 µg/mL glycogen); the buffer-only well is NEG-glyc.

1.2 Important note(s)

  1. Units. Promega’s glycogen standards and linear range are specified in µg/mL (linear to 20 µg/mL), and the layout labels (STD-glyc-20) are on that scale. Earlier scripts in this repo labeled the same quantity “µg/µL”; that label is incorrect by a factor of 1000 and is corrected here.

  2. Tissue weights are taken from the plate layout labels rather than a separate weights CSV, since they are already encoded there for this experiment.

2 SETUP

2.1 Libraries

library(knitr)
library(ggplot2)
library(dplyr)
## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union
library(tidyr)

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,         # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",        # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Output directory

# Output directory for this analysis (matches this file's name, per ../code/README.md)
output_dir <- "../output/Gen5-20260730-mgig-fams1_9-glycogen-glo"
dir.create(output_dir, showWarnings = FALSE, recursive = TRUE)

cat("--- output_dir: destination for all figures and tables ---\n")
str(output_dir)
--- output_dir: destination for all figures and tables ---
 chr "../output/Gen5-20260730-mgig-fams1_9-glycogen-glo"

3 DATA

Data are read from the local repo (../data/raw_luminescence/) so this document renders before/after the files are pushed to GitHub. The commented-out URLs are the remote equivalents.

data_dir <- "../data/raw_luminescence"

# plate 1 - all 32 samples, triplicate, 1:25 dilution

plate_layout1 <- read.csv(file.path(data_dir, "layout-Gen5-20260730-mgig-fams1_9-plate-01.csv"),
                          header = FALSE, stringsAsFactors = FALSE)
raw_luminescence1 <- read.csv(file.path(data_dir, "raw_lum-Gen5-20260730-mgig-fams1_9-plate-01.csv"),
                              header = FALSE)

# plate 2 - glycogen standard curve + negative control, triplicates
plate_layout2 <- read.csv(file.path(data_dir, "layout-Gen5-20260730-mgig-fams1_9-plate-02.csv"),
                          header = FALSE, stringsAsFactors = FALSE)
raw_luminescence2 <- read.csv(file.path(data_dir, "raw_lum-Gen5-20260730-mgig-fams1_9-plate-02.csv"),
                              header = FALSE)

cat("Plate layout 1 (samples):\n\n")
str(plate_layout1)
cat("\nRaw luminescence 1 (samples):\n\n")
str(raw_luminescence1)
cat("\nPlate layout 2 (standards):\n\n")
str(plate_layout2)
cat("\nRaw luminescence 2 (standards):\n\n")
str(raw_luminescence2)
Plate layout 1 (samples):

'data.frame':   8 obs. of  12 variables:
 $ V1 : chr  "1_01_ambient-glyc-7.0-df.25" "1_05_ambient-glyc-7.9-df.25" "1_01_36C-glyc-9.6-df.25" "1_05_36C-glyc-25.7-df.25" ...
 $ V2 : chr  "1_01_ambient-glyc-7.0-df.25" "1_05_ambient-glyc-7.9-df.25" "1_01_36C-glyc-9.6-df.25" "1_05_36C-glyc-25.7-df.25" ...
 $ V3 : chr  "1_01_ambient-glyc-7.0-df.25" "1_05_ambient-glyc-7.9-df.25" "1_01_36C-glyc-9.6-df.25" "1_05_36C-glyc-25.7-df.25" ...
 $ V4 : chr  "1_02_ambient-glyc-4.7-df.25" "1_06_ambient-glyc-11.0-df.25" "1_02_36C-glyc-9.2-df.25" "1_06_36C-glyc-13.0-df.25" ...
 $ V5 : chr  "1_02_ambient-glyc-4.7-df.25" "1_06_ambient-glyc-11.0-df.25" "1_02_36C-glyc-9.2-df.25" "1_06_36C-glyc-13.0-df.25" ...
 $ V6 : chr  "1_02_ambient-glyc-4.7-df.25" "1_06_ambient-glyc-11.0-df.25" "1_02_36C-glyc-9.2-df.25" "1_06_36C-glyc-13.0-df.25" ...
 $ V7 : chr  "1_03_ambient-glyc-2.5-df.25" "1_07_ambient-glyc-6.1-df.25" "1_03_36C-glyc-3.1-df.25" "1_07_36C-glyc-6.1-df.25" ...
 $ V8 : chr  "1_03_ambient-glyc-2.5-df.25" "1_07_ambient-glyc-6.1-df.25" "1_03_36C-glyc-3.1-df.25" "1_07_36C-glyc-6.1-df.25" ...
 $ V9 : chr  "1_03_ambient-glyc-2.5-df.25" "1_07_ambient-glyc-6.1-df.25" "1_03_36C-glyc-3.1-df.25" "1_07_36C-glyc-6.1-df.25" ...
 $ V10: chr  "1_04_ambient-glyc-10.7-df.25" "1_08_ambient-glyc-24.8-df.25" "1_04_36C-glyc-12.4-df.25" "1_08_36C-glyc-8.8-df.25" ...
 $ V11: chr  "1_04_ambient-glyc-10.7-df.25" "1_08_ambient-glyc-24.8-df.25" "1_04_36C-glyc-12.4-df.25" "1_08_36C-glyc-8.8-df.25" ...
 $ V12: chr  "1_04_ambient-glyc-10.7-df.25" "1_08_ambient-glyc-24.8-df.25" "1_04_36C-glyc-12.4-df.25" "1_08_36C-glyc-8.8-df.25" ...

Raw luminescence 1 (samples):

'data.frame':   8 obs. of  12 variables:
 $ V1 : int  11367 21917 32434 99029 7587 70560 15396 47510
 $ V2 : int  10284 22772 33069 89772 7324 70742 13953 51778
 $ V3 : int  10754 21609 30029 93511 7098 69326 13889 50054
 $ V4 : int  14834 99291 5244 56801 4924 3792 11887 13583
 $ V5 : int  14603 101901 5895 55712 4796 3376 12805 14333
 $ V6 : int  14460 90681 5898 57887 5252 3778 12878 14402
 $ V7 : int  7584 5141 4542 22179 10788 10100 44070 2685
 $ V8 : int  7058 4776 5385 20629 11788 9429 45147 2382
 $ V9 : int  7476 5310 4568 22171 11825 10378 42353 2157
 $ V10: int  39832 200695 6309 46456 5276 2669 23819 19314
 $ V11: int  36805 200780 3369 47737 5688 2825 23562 17222
 $ V12: int  41820 201408 5548 44512 5994 2539 24569 19417

Plate layout 2 (standards):

'data.frame':   8 obs. of  12 variables:
 $ V1 : chr  "STD-glyc-20" "STD-glyc-0" "NEG-glyc" "" ...
 $ V2 : chr  "STD-glyc-20" "STD-glyc-0" "NEG-glyc" "" ...
 $ V3 : chr  "STD-glyc-20" "STD-glyc-0" "NEG-glyc" "" ...
 $ V4 : chr  "STD-glyc-2" "" "" "" ...
 $ V5 : chr  "STD-glyc-2" "" "" "" ...
 $ V6 : chr  "STD-glyc-2" "" "" "" ...
 $ V7 : chr  "STD-glyc-0.2" "" "" "" ...
 $ V8 : chr  "STD-glyc-0.2" "" "" "" ...
 $ V9 : chr  "STD-glyc-0.2" "" "" "" ...
 $ V10: chr  "STD-glyc-0.02" "" "" "" ...
 $ V11: chr  "STD-glyc-0.02" "" "" "" ...
 $ V12: chr  "STD-glyc-0.02" "" "" "" ...

Raw luminescence 2 (standards):

'data.frame':   8 obs. of  12 variables:
 $ V1 : int  53166 924 932 NA NA NA NA NA
 $ V2 : int  68390 909 840 NA NA NA NA NA
 $ V3 : int  63329 897 967 NA NA NA NA NA
 $ V4 : int  6947 NA NA NA NA NA NA NA
 $ V5 : int  6432 NA NA NA NA NA NA NA
 $ V6 : int  6156 NA NA NA NA NA NA NA
 $ V7 : int  1437 NA NA NA NA NA NA NA
 $ V8 : int  1415 NA NA NA NA NA NA NA
 $ V9 : int  1377 NA NA NA NA NA NA NA
 $ V10: int  1144 NA NA NA NA NA NA NA
 $ V11: int  1153 NA NA NA NA NA NA NA
 $ V12: int  963 NA NA NA NA NA NA NA

3.1 Reshape plates to long format

Rather than defining one variable per sample, both plates are melted to a tidy well-level data frame and the sample metadata is parsed out of the layout labels. This keeps the code independent of how many samples are on the plate.

# Convert a paired layout/luminescence plate into one row per well
plate_to_long <- function(layout, luminescence, plate_label) {
  n_row <- 8
  n_col <- 12
  out <- expand.grid(plate_row_idx = 1:n_row, plate_col = 1:n_col)
  out$plate       <- plate_label
  out$plate_row   <- LETTERS[out$plate_row_idx]
  out$well        <- sprintf("%s%02d", out$plate_row, out$plate_col)
  out$label       <- trimws(as.character(mapply(function(i, j) layout[i, j],
                                                out$plate_row_idx, out$plate_col)))
  out$luminescence <- as.numeric(mapply(function(i, j) luminescence[i, j],
                                        out$plate_row_idx, out$plate_col))
  out <- out[order(out$plate_row_idx, out$plate_col), ]
  out[out$label != "" & !is.na(out$label), ]
}

plate1_long <- plate_to_long(plate_layout1, raw_luminescence1, "plate-01")
plate2_long <- plate_to_long(plate_layout2, raw_luminescence2, "plate-02")


cat("--- plate_to_long(): layout + luminescence -> one row per well ---\n\n")
str(args(plate_to_long))

cat("\n--- plate1_long: one row per occupied well, sample plate ---\n\n")
str(plate1_long)

cat("\n--- plate2_long: one row per occupied well, standards plate ---\n\n")
str(plate2_long)
--- plate_to_long(): layout + luminescence -> one row per well ---

function (layout, luminescence, plate_label)  

--- plate1_long: one row per occupied well, sample plate ---

'data.frame':   96 obs. of  7 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "plate-01" "plate-01" "plate-01" "plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_02_ambient-glyc-4.7-df.25" ...
 $ luminescence : num  11367 10284 10754 14834 14603 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- plate2_long: one row per occupied well, standards plate ---

'data.frame':   18 obs. of  7 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "plate-02" "plate-02" "plate-02" "plate-02" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "STD-glyc-20" "STD-glyc-20" "STD-glyc-20" "STD-glyc-2" ...
 $ luminescence : num  53166 68390 63329 6947 6432 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

4 STANDARD CURVES

4.1 Glycogen Standard Curve

4.1.1 Extract luminescence data

Plate 02, row A holds the four non-zero glycogen standards (20, 2, 0.2, 0.02 µg/mL) in triplicate; row B columns 1-3 hold the STD-glyc-0 (zero-glycogen) standard and row C columns 1-3 hold the NEG-glyc buffer-only negative control.

standards <- plate2_long %>%
  filter(grepl("^STD-glyc-", label)) %>%
  mutate(glyc_concentration = as.numeric(sub("^STD-glyc-", "", label)))

negative_control <- plate2_long %>% filter(grepl("^NEG-glyc", label))

cat("Standard concentrations (ug/mL):",
    paste(sort(unique(standards$glyc_concentration)), collapse = ", "), "\n")
cat("Replicates per standard:",
    paste(unique(table(standards$glyc_concentration)), collapse = ", "), "\n\n")

cat("Negative control (buffer only) luminescence:",
    paste(negative_control$luminescence, collapse = ", "),
    "| mean =", round(mean(negative_control$luminescence), 1), "\n")
cat("Zero glycogen standard luminescence:  mean =",
    round(mean(standards$luminescence[standards$glyc_concentration == 0]), 1), "\n")

cat("\n--- standards: standard curve wells with parsed concentration ---\n")
str(standards)
cat("\n--- negative_control: buffer-only wells ---\n")
str(negative_control)
Standard concentrations (ug/mL): 0, 0.02, 0.2, 2, 20 
Replicates per standard: 3 

Negative control (buffer only) luminescence: 932, 840, 967 | mean = 913 
Zero glycogen standard luminescence:  mean = 910 

--- standards: standard curve wells with parsed concentration ---
'data.frame':   15 obs. of  8 variables:
 $ plate_row_idx     : int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col         : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate             : chr  "plate-02" "plate-02" "plate-02" "plate-02" ...
 $ plate_row         : chr  "A" "A" "A" "A" ...
 $ well              : chr  "A01" "A02" "A03" "A04" ...
 $ label             : chr  "STD-glyc-20" "STD-glyc-20" "STD-glyc-20" "STD-glyc-2" ...
 $ luminescence      : num  53166 68390 63329 6947 6432 ...
 $ glyc_concentration: num  20 20 20 2 2 2 0.2 0.2 0.2 0.02 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

--- negative_control: buffer-only wells ---
'data.frame':   3 obs. of  7 variables:
 $ plate_row_idx: int  3 3 3
 $ plate_col    : int  1 2 3
 $ plate        : chr  "plate-02" "plate-02" "plate-02"
 $ plate_row    : chr  "C" "C" "C"
 $ well         : chr  "C01" "C02" "C03"
 $ label        : chr  "NEG-glyc" "NEG-glyc" "NEG-glyc"
 $ luminescence : num  932 840 967
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

The buffer-only negative control and the zero-glycogen standard are indistinguishable, as expected – both report the reagent/instrument background.

4.1.2 Glycogen standard curve summary statistics and linear regression

Following the convention used in previous analyses in this repo, the regression is fit to the mean luminescence at each standard concentration. The fit to all individual replicate wells is also reported, since it is the more honest estimate of the curve’s uncertainty.

glycogen_summary_data <- standards %>%
  group_by(glyc_concentration) %>%
  summarise(
    glyc_mean_luminescence = mean(luminescence),
    glyc_sd                = sd(luminescence),
    glyc_se                = sd(luminescence) / sqrt(n()),
    glyc_cv                = 100 * sd(luminescence) / mean(luminescence),
    glyc_n                 = n(),
    .groups = "drop"
  ) %>%
  arrange(glyc_concentration)

# Regression on the per-concentration means (repo convention)
lm_model      <- lm(glyc_mean_luminescence ~ glyc_concentration, data = glycogen_summary_data)
glyc_slope     <- coef(lm_model)[2]
glyc_intercept <- coef(lm_model)[1]
glyc_r_squared <- summary(lm_model)$r.squared

# Regression on all individual replicate wells (for comparison)
lm_model_reps    <- lm(luminescence ~ glyc_concentration, data = standards)
glyc_r2_reps     <- summary(lm_model_reps)$r.squared

glyc_conc_min_nonzero <- min(glycogen_summary_data$glyc_concentration[
  glycogen_summary_data$glyc_concentration > 0])
glyc_conc_max <- max(glycogen_summary_data$glyc_concentration)

kable(glycogen_summary_data,
      digits = c(2, 1, 1, 1, 2, 0),
      col.names = c("Glycogen (ug/mL)", "Mean luminescence", "SD", "SEM", "CV (%)", "n"),
      caption = "Glycogen standard curve summary statistics")

cat("\nFit to concentration means:      y =", sprintf("%.1f", glyc_slope), "x +",
    sprintf("%.1f", glyc_intercept), " R^2 =", sprintf("%.4f", glyc_r_squared), "\n")
cat("Fit to individual replicates:    R^2 =", sprintf("%.4f", glyc_r2_reps), "\n")
cat("Quantifiable range (standards):", glyc_conc_min_nonzero, "-", glyc_conc_max, "ug/mL\n")

cat("\n--- glycogen_summary_data: per-concentration summary statistics ---\n")
str(glycogen_summary_data)
cat("\n--- lm_model: regression on per-concentration means ---\n")
str(lm_model, max.level = 1, give.attr = FALSE)
cat("\n--- lm_model_reps: regression on all individual replicate wells ---\n")
str(lm_model_reps, max.level = 1, give.attr = FALSE)
Glycogen (ug/mL) Mean luminescence SD SEM CV (%) n
0.00 910.0 13.5 7.8 1.49 3
0.02 1086.7 107.2 61.9 9.86 3
0.20 1409.7 30.4 17.5 2.15 3
2.00 6511.7 401.5 231.8 6.17 3
20.00 61628.3 7753.2 4476.3 12.58 3

Glycogen standard curve summary statistics

Fit to concentration means:      y = 3039.5 x + 801.8  R^2 = 0.9999 
Fit to individual replicates:    R^2 = 0.9859 
Quantifiable range (standards): 0.02 - 20 ug/mL

--- glycogen_summary_data: per-concentration summary statistics ---
tibble [5 × 6] (S3: tbl_df/tbl/data.frame)
 $ glyc_concentration    : num [1:5] 0 0.02 0.2 2 20
 $ glyc_mean_luminescence: num [1:5] 910 1087 1410 6512 61628
 $ glyc_sd               : num [1:5] 13.5 107.2 30.4 401.5 7753.2
 $ glyc_se               : num [1:5] 7.81 61.89 17.52 231.79 4476.3
 $ glyc_cv               : num [1:5] 1.49 9.86 2.15 6.17 12.58
 $ glyc_n                : int [1:5] 3 3 3 3 3

--- lm_model: regression on per-concentration means ---
List of 12
 $ coefficients : Named num [1:2] 802 3039
 $ residuals    : Named num [1:5] 108.2489 224.1257 0.0166 -369.0743 36.6831
 $ effects      : Named num [1:5] -31996.5 53108.1 -60.1 -412.5 160.5
 $ rank         : int 2
 $ fitted.values: Named num [1:5] 802 863 1410 6881 61592
 $ assign       : int [1:2] 0 1
 $ qr           :List of 5
 $ df.residual  : int 3
 $ xlevels      : Named list()
 $ call         : language lm(formula = glyc_mean_luminescence ~ glyc_concentration, data = glycogen_summary_data)
 $ terms        :Classes 'terms', 'formula'  language glyc_mean_luminescence ~ glyc_concentration
 $ model        :'data.frame':  5 obs. of  2 variables:

--- lm_model_reps: regression on all individual replicate wells ---
List of 12
 $ coefficients : Named num [1:2] 802 3039
 $ residuals    : Named num [1:15] -8425.7 6798.3 1737.3 66.3 -448.7 ...
 $ effects      : Named num [1:15] -55420 -91986 993 2923 2408 ...
 $ rank         : int 2
 $ fitted.values: Named num [1:15] 61592 61592 61592 6881 6881 ...
 $ assign       : int [1:2] 0 1
 $ qr           :List of 5
 $ df.residual  : int 13
 $ xlevels      : Named list()
 $ call         : language lm(formula = luminescence ~ glyc_concentration, data = standards)
 $ terms        :Classes 'terms', 'formula'  language luminescence ~ glyc_concentration
 $ model        :'data.frame':  15 obs. of  2 variables:

4.1.3 Extract sample data and calculate glycogen levels

Sample labels are parsed into family, individual, temperature, tissue weight, and dilution factor. The delimiter inconsistencies noted above (1_02-36C, 9-07-36C, 9-08-36C) are normalized here.

samples_wells <- plate1_long %>%
  # Normalize `-` to `_` in the <family>-<individual>-<temperature> prefix
  mutate(label_clean = sub("^([19])[-_]([0-9]{2})[-_](ambient|36C)-", "\\1_\\2_\\3-", label)) %>%
  mutate(
    sample_id  = sub("-glyc-.*$", "", label_clean),
    family     = sub("^([19])_.*$", "\\1", label_clean),
    individual = sub("^[19]_([0-9]{2})_.*$", "\\1", label_clean),
    temperature = sub("^[19]_[0-9]{2}_([^-]+)-.*$", "\\1", label_clean),
    weight_mg  = as.numeric(sub("^.*-glyc-([0-9.]+)-df\\..*$", "\\1", label_clean)),
    dilution   = as.numeric(sub("^.*-df\\.", "", label_clean))
  ) %>%
  mutate(
    family      = factor(paste("Family", family), levels = c("Family 1", "Family 9")),
    temperature = factor(temperature, levels = c("ambient", "36C"))
  )

# Fail loudly rather than silently dropping malformed labels
stopifnot(!any(is.na(samples_wells$weight_mg)),
          !any(is.na(samples_wells$dilution)),
          !any(is.na(samples_wells$temperature)))

cat("Wells parsed:", nrow(samples_wells),
    "| unique samples:", length(unique(samples_wells$sample_id)), "\n")
cat("Dilution factor(s) used:", paste(unique(samples_wells$dilution), collapse = ", "), "\n\n")
cat("Design (wells per group):\n")
print(table(samples_wells$family, samples_wells$temperature))
cat("\nTissue weight range (mg):",
    paste(range(samples_wells$weight_mg), collapse = " - "), "\n")

cat("\n--- samples_wells: one row per sample well, metadata parsed from label ---\n")
str(samples_wells)
Wells parsed: 96 | unique samples: 32 
Dilution factor(s) used: 25 

Design (wells per group):
          
           ambient 36C
  Family 1      24  24
  Family 9      24  24

Tissue weight range (mg): 1.8 - 25.7 

--- samples_wells: one row per sample well, metadata parsed from label ---
'data.frame':   96 obs. of  14 variables:
 $ plate_row_idx: int  1 1 1 1 1 1 1 1 1 1 ...
 $ plate_col    : int  1 2 3 4 5 6 7 8 9 10 ...
 $ plate        : chr  "plate-01" "plate-01" "plate-01" "plate-01" ...
 $ plate_row    : chr  "A" "A" "A" "A" ...
 $ well         : chr  "A01" "A02" "A03" "A04" ...
 $ label        : chr  "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_02_ambient-glyc-4.7-df.25" ...
 $ luminescence : num  11367 10284 10754 14834 14603 ...
 $ label_clean  : chr  "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_01_ambient-glyc-7.0-df.25" "1_02_ambient-glyc-4.7-df.25" ...
 $ sample_id    : chr  "1_01_ambient" "1_01_ambient" "1_01_ambient" "1_02_ambient" ...
 $ family       : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual   : chr  "01" "01" "01" "02" ...
 $ temperature  : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 1 1 ...
 $ weight_mg    : num  7 7 7 4.7 4.7 4.7 2.5 2.5 2.5 10.7 ...
 $ dilution     : num  25 25 25 25 25 25 25 25 25 25 ...
 - attr(*, "out.attrs")=List of 2
  ..$ dim     : Named int [1:2] 8 12
  .. ..- attr(*, "names")= chr [1:2] "plate_row_idx" "plate_col"
  ..$ dimnames:List of 2
  .. ..$ plate_row_idx: chr [1:8] "plate_row_idx=1" "plate_row_idx=2" "plate_row_idx=3" "plate_row_idx=4" ...
  .. ..$ plate_col    : chr [1:12] "plate_col= 1" "plate_col= 2" "plate_col= 3" "plate_col= 4" ...

Per-sample values are then computed by averaging the technical triplicate and back-calculating through the standard curve:

  1. Well concentration = (mean luminescence - intercept) / slope → µg/mL in the assayed (diluted) well.
  2. Homogenate concentration = well concentration × dilution factor (25) → µg/mL in the undiluted homogenate.
  3. Normalized glycogen = homogenate concentration / tissue weight → µg/mL per mg tissue.

Note on step 3: because homogenization volume is not recorded in these data files, normalized values are reported as µg glycogen per mL homogenate per mg tissue. If the standard 1.0 mL homogenization volume was used (750 µL PBS/0.3N HCl + 250 µL Tris, per ../glycogen-promega-notes.md), these values are numerically equal to µg glycogen / mg tissue. All comparisons below are unaffected as long as the volume was constant across samples.

sample_glycogen <- samples_wells %>%
  group_by(sample_id, family, individual, temperature, weight_mg, dilution) %>%
  summarise(
    n_reps        = n(),
    mean_lum      = mean(luminescence),
    sd_lum        = sd(luminescence),
    cv_lum        = 100 * sd(luminescence) / mean(luminescence),
    .groups = "drop"
  ) %>%
  mutate(
    well_conc_ug_mL       = (mean_lum - glyc_intercept) / glyc_slope,
    homogenate_conc_ug_mL = well_conc_ug_mL * dilution,
    norm_glycogen         = homogenate_conc_ug_mL / weight_mg,
    in_std_range          = well_conc_ug_mL >= glyc_conc_min_nonzero &
                            well_conc_ug_mL <= glyc_conc_max
  ) %>%
  arrange(family, temperature, individual)

cat("Samples quantified:", nrow(sample_glycogen), "\n")

cat("\n--- sample_glycogen: one row per individual, triplicate averaged and back-calculated ---\n")
str(sample_glycogen)
Samples quantified: 32 

--- sample_glycogen: one row per individual, triplicate averaged and back-calculated ---
tibble [32 × 14] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:32] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:32] "01" "02" "03" "04" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 2 2 ...
 $ weight_mg            : num [1:32] 7 4.7 2.5 10.7 7.9 11 6.1 24.8 9.6 9.2 ...
 $ dilution             : num [1:32] 25 25 25 25 25 25 25 25 25 25 ...
 $ n_reps               : int [1:32] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:32] 10802 14632 7373 39486 22099 ...
 $ sd_lum               : num [1:32] 543 189 278 2525 603 ...
 $ cv_lum               : num [1:32] 5.03 1.29 3.77 6.4 2.73 ...
 $ well_conc_ug_mL      : num [1:32] 3.29 4.55 2.16 12.73 7.01 ...
 $ homogenate_conc_ug_mL: num [1:32] 82.2 113.8 54 318.2 175.2 ...
 $ norm_glycogen        : num [1:32] 11.7 24.2 21.6 29.7 22.2 ...
 $ in_std_range         : logi [1:32] TRUE TRUE TRUE TRUE TRUE FALSE ...

4.1.4 Plot glycogen standard curve with sample points

std_curve_plot <- ggplot(glycogen_summary_data,
                         aes(x = glyc_concentration, y = glyc_mean_luminescence)) +
  geom_smooth(aes(linetype = "Std curve best fit line"), method = "lm", se = FALSE,
              color = "coral", linewidth = 1) +
  geom_errorbar(aes(ymin = glyc_mean_luminescence - glyc_se,
                    ymax = glyc_mean_luminescence + glyc_se),
                width = 0.3, linewidth = 1, color = "darkblue") +
  geom_point(aes(color = "Glycogen standard"), shape = 16, size = 4) +
  geom_point(data = sample_glycogen,
             aes(x = well_conc_ug_mL, y = mean_lum, color = family),
             shape = 17, size = 2.5, alpha = 0.85) +
  geom_point(data = filter(sample_glycogen, !in_std_range),
             aes(x = well_conc_ug_mL, y = mean_lum, shape = "Extrapolated"),
             size = 4, stroke = 1.1, color = "red", fill = NA) +
  annotate("rect", xmin = glyc_conc_max, xmax = Inf, ymin = -Inf, ymax = Inf,
           fill = "red", alpha = 0.06) +
  scale_color_manual(name = "",
                     breaks = c("Glycogen standard", "Family 1", "Family 9"),
                     values = c("Glycogen standard" = "steelblue",
                                "Family 1"          = "#1b7837",
                                "Family 9"          = "#762a83")) +
  scale_shape_manual(name = "", values = c("Extrapolated" = 21)) +
  scale_linetype_manual(name = "", values = c("Std curve best fit line" = "dashed")) +
  guides(color = guide_legend(override.aes = list(shape = c(16, 17, 17), size = 3),
                              order = 1),
         shape = guide_legend(order = 2), linetype = guide_legend(order = 3)) +
  annotate("label",
           x = glyc_conc_max * 0.45, y = max(glycogen_summary_data$glyc_mean_luminescence) * 0.95,
           label = sprintf("y = %.1fx + %.1f\nR2 = %.4f", glyc_slope, glyc_intercept, glyc_r_squared),
           size = 3.5, fontface = "bold", fill = "white", color = "coral",
           label.padding = unit(0.3, "lines")) +
  labs(
    title = "Glycogen standard curve with sample well concentrations",
    subtitle = "Shaded region is above the top standard (20 ug/mL): samples there are extrapolated",
    x = "Glycogen concentration (ug/mL)",
    y = "Luminescence",
    caption = "Standard curve error bars are SEM of triplicate wells (n = 3)"
  ) +
  theme_bw() +
  theme(
    plot.title  = element_text(size = 14, face = "bold"),
    axis.title  = element_text(size = 12, face = "bold"),
    panel.grid.minor = element_line(linetype = "dashed", color = "grey70")
  )

cat("--- std_curve_plot: ggplot object structure ---\n")
summary(std_curve_plot)

ggsave(file.path(output_dir, "glycogen_standard_curve.png"), std_curve_plot,
       width = 10, height = 6, dpi = 300)

std_curve_plot

--- std_curve_plot: ggplot object structure ---
data: glyc_concentration, glyc_mean_luminescence, glyc_sd, glyc_se,
  glyc_cv, glyc_n [5x6]
mapping:  x = ~glyc_concentration, y = ~glyc_mean_luminescence
scales:   colour, shape, linetype 
faceting:  <empty> 
-----------------------------------
mapping: linetype = Std curve best fit line 
geom_smooth: na.rm = FALSE, orientation = NA, se = FALSE
stat_smooth: na.rm = FALSE, orientation = NA, se = FALSE, method = lm
position_identity 

mapping: ymin = ~glyc_mean_luminescence - glyc_se, ymax = ~glyc_mean_luminescence + glyc_se 
geom_errorbar: na.rm = FALSE, orientation = NA, lineend = butt, width = 0.3
stat_identity: na.rm = FALSE
position_identity 

mapping: colour = Glycogen standard 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: x = ~well_conc_ug_mL, y = ~mean_lum, colour = ~family 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: x = ~well_conc_ug_mL, y = ~mean_lum, shape = Extrapolated 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: xmin = ~xmin, xmax = ~xmax, ymin = ~ymin, ymax = ~ymax 
geom_rect: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

mapping: x = ~x, y = ~y 
geom_label: na.rm = FALSE, label.padding = 0.3
stat_identity: na.rm = FALSE
position_identity

The linear fit is anchored almost entirely by the 20 µg/mL standard. On a log-log scale it is clear that the three lowest standards (0.02, 0.2, 2 µg/mL) sit essentially on the baseline of this linear model, i.e. the curve has little resolving power at the bottom of the range:

std_curve_log_plot <- ggplot(filter(glycogen_summary_data, glyc_concentration > 0),
                             aes(x = glyc_concentration, y = glyc_mean_luminescence)) +
  geom_hline(aes(yintercept = mean(negative_control$luminescence),
                 linetype = "Background (buffer only)"), color = "grey40") +
  geom_line(color = "steelblue", linewidth = 0.8) +
  geom_errorbar(aes(ymin = glyc_mean_luminescence - glyc_se,
                    ymax = glyc_mean_luminescence + glyc_se),
                width = 0.08, color = "darkblue") +
  geom_point(color = "steelblue", size = 3.5) +
  scale_x_log10() +
  scale_y_log10() +
  scale_linetype_manual(name = "", values = c("Background (buffer only)" = "dotted")) +
  labs(
    title = "Glycogen standard curve (log-log)",
    x = "Glycogen concentration (ug/mL, log scale)",
    y = "Luminescence (log scale)"
  ) +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.title = element_text(size = 11, face = "bold"))

cat("--- std_curve_log_plot: ggplot object structure ---\n")
summary(std_curve_log_plot)

ggsave(file.path(output_dir, "glycogen_standard_curve_loglog.png"), std_curve_log_plot,
       width = 8, height = 5.5, dpi = 300)

std_curve_log_plot

--- std_curve_log_plot: ggplot object structure ---
data: glyc_concentration, glyc_mean_luminescence, glyc_sd, glyc_se,
  glyc_cv, glyc_n [4x6]
mapping:  x = ~glyc_concentration, y = ~glyc_mean_luminescence
scales:   x, xmin, xmax, xend, xintercept, xmin_final, xmax_final, xlower, xmiddle, xupper, x0, y, ymin, ymax, yend, yintercept, ymin_final, ymax_final, lower, middle, upper, y0, linetype 
faceting:  <empty> 
-----------------------------------
mapping: yintercept = ~mean(negative_control$luminescence), linetype = Background (buffer only) 
geom_hline: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: ymin = ~glyc_mean_luminescence - glyc_se, ymax = ~glyc_mean_luminescence + glyc_se 
geom_errorbar: na.rm = FALSE, orientation = NA, lineend = butt, width = 0.08
stat_identity: na.rm = FALSE
position_identity 

geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

5 QUALITY CONTROL

5.1 Technical replicate variability

cat("Coefficient of variation across technical triplicates (%):\n")
print(round(summary(sample_glycogen$cv_lum), 2))

cat("\nSamples with CV > 15%:\n")
high_cv <- sample_glycogen %>% filter(cv_lum > 15) %>%
  select(sample_id, mean_lum, sd_lum, cv_lum)
if (nrow(high_cv) == 0) {
  cat("  none\n")
} else {
  print(as.data.frame(high_cv), row.names = FALSE, digits = 4)
}

cat("\n--- high_cv: samples exceeding the 15% triplicate CV threshold ---\n\n")
str(high_cv)
Coefficient of variation across technical triplicates (%):
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
   0.19    3.31    4.92    5.48    6.12   30.07 

Samples with CV > 15%:
 sample_id mean_lum sd_lum cv_lum
  1_04_36C     5075   1526  30.07

--- high_cv: samples exceeding the 15% triplicate CV threshold ---

tibble [1 × 4] (S3: tbl_df/tbl/data.frame)
 $ sample_id: chr "1_04_36C"
 $ mean_lum : num 5075
 $ sd_lum   : num 1526
 $ cv_lum   : num 30.1

5.2 Samples outside the standard curve range

oor <- sample_glycogen %>%
  filter(!in_std_range) %>%
  select(sample_id, family, temperature, mean_lum, well_conc_ug_mL,
         weight_mg, norm_glycogen) %>%
  arrange(desc(well_conc_ug_mL))

cat("Samples with well concentrations outside", glyc_conc_min_nonzero, "-",
    glyc_conc_max, "ug/mL:", nrow(oor), "of", nrow(sample_glycogen), "\n\n")
kable(oor, digits = c(0, 0, 0, 0, 2, 1, 2),
      col.names = c("Sample", "Family", "Temp", "Mean lum.", "Well conc. (ug/mL)",
                    "Tissue (mg)", "Normalized glycogen"),
      caption = "Samples requiring extrapolation beyond the top standard")

cat("\n--- Samples outside the standard curve range ---\n\n")
str(oor)
Samples with well concentrations outside 0.02 - 20 ug/mL: 4 of 32
Sample Family Temp Mean lum. Well conc. (ug/mL) Tissue (mg) Normalized glycogen
1_08_ambient Family 1 ambient 200961 65.85 24.8 66.38
1_06_ambient Family 1 ambient 97291 31.75 11.0 72.15
1_05_36C Family 1 36C 94104 30.70 25.7 29.86
9_05_ambient Family 9 ambient 70209 22.84 9.2 62.05

Samples requiring extrapolation beyond the top standard

--- Samples outside the standard curve range ---

tibble [4 × 7] (S3: tbl_df/tbl/data.frame)
 $ sample_id      : chr [1:4] "1_08_ambient" "1_06_ambient" "1_05_36C" "9_05_ambient"
 $ family         : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 2
 $ temperature    : Factor w/ 2 levels "ambient","36C": 1 1 2 1
 $ mean_lum       : num [1:4] 200961 97291 94104 70209
 $ well_conc_ug_mL: num [1:4] 65.9 31.7 30.7 22.8
 $ weight_mg      : num [1:4] 24.8 11 25.7 9.2
 $ norm_glycogen  : num [1:4] 66.4 72.1 29.9 62.1

All four are above the 20 µg/mL top standard, so their concentrations are extrapolated rather than interpolated. Notably, 1_08_ambient reads ~201,000 RLU (3.3× the top standard) with a very tight triplicate CV (0.2%), which is the signature of a well near the top of the detector’s usable range. The other three (1_06_ambient, 1_05_36C, 9_05_ambient) are more modestly above range (~70,000-97,000 RLU, roughly 1.5-1.6× the top standard). These samples should be re-run at a higher dilution (e.g. 1:50-1:100 for 1_08_ambient) before the values are treated as quantitative. Group comparisons below are therefore repeated with these samples excluded as a sensitivity check.

5.3 Plate 01 luminescence map

A well-position map checks for edge effects or systematic gradients that would confound the family/temperature contrasts. In this design, family and temperature are both encoded as two-row blocks (rows A/B = family 1 ambient, C/D = family 1 36°C, E/F = family 9 ambient, G/H = family 9 36°C), while column position encodes only the individual (with each individual’s triplicate occupying 3 adjacent columns). Any row-wise artifact would therefore alias directly onto the family/temperature contrasts of interest; a column-wise artifact would not, since individuals are distributed independently of family and temperature group.

plate_map_plot <- ggplot(samples_wells,
                         aes(x = factor(plate_col), y = factor(plate_row, levels = rev(LETTERS[1:8])))) +
  geom_tile(aes(fill = log10(luminescence)), color = "white", linewidth = 0.5) +
  geom_text(aes(label = format(luminescence, big.mark = ",")), size = 2.4, color = "grey15") +
  scale_fill_viridis_c(option = "magma", name = "log10(RLU)") +
  labs(
    title = "Plate 01 raw luminescence by well",
    subtitle = "Row-pairs A/B, C/D, E/F, G/H = family1-amb, family1-36C, family9-amb, family9-36C; columns = individual (triplicate)",
    x = "Plate column", y = "Plate row"
  ) +
  theme_minimal() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        panel.grid = element_blank())

cat("--- plate_map_plot: ggplot object structure ---\n\n")
summary(plate_map_plot)

ggsave(file.path(output_dir, "plate01_luminescence_map.png"), plate_map_plot,
       width = 10, height = 5, dpi = 300)

plate_map_plot

--- plate_map_plot: ggplot object structure ---

data: plate_row_idx, plate_col, plate, plate_row, well, label,
  luminescence, label_clean, sample_id, family, individual,
  temperature, weight_mg, dilution [96x14]
mapping:  x = ~factor(plate_col), y = ~factor(plate_row, levels = rev(LETTERS[1:8]))
scales:   fill 
faceting:  <empty> 
-----------------------------------
mapping: fill = ~log10(luminescence) 
geom_tile: na.rm = FALSE, lineend = butt, linejoin = mitre
stat_identity: na.rm = FALSE
position_identity 

mapping: label = ~format(luminescence, big.mark = ",") 
geom_text: na.rm = FALSE, parse = FALSE, check_overlap = FALSE, size.unit = mm
stat_identity: na.rm = FALSE
position_nudge

5.4 Relationship between tissue weight and signal

If normalization by tissue weight is working, normalized glycogen should show no strong trend against input tissue weight.

weight_check <- sample_glycogen %>%
  select(sample_id, family, temperature, weight_mg,
         homogenate_conc_ug_mL, norm_glycogen) %>%
  pivot_longer(c(homogenate_conc_ug_mL, norm_glycogen),
               names_to = "metric", values_to = "value") %>%
  mutate(metric = recode(metric,
                         homogenate_conc_ug_mL = "Homogenate glycogen (ug/mL)",
                         norm_glycogen         = "Normalized glycogen (ug/mL/mg tissue)"))

weight_plot <- ggplot(weight_check, aes(x = weight_mg, y = value)) +
  geom_smooth(method = "lm", se = FALSE, color = "grey50", linetype = "dashed") +
  geom_point(aes(color = family, shape = temperature), size = 2.8) +
  facet_wrap(~ metric, scales = "free_y") +
  scale_color_manual(values = c("Family 1" = "#1b7837", "Family 9" = "#762a83"),
                     name = "") +
  scale_shape_manual(values = c("ambient" = 16, "36C" = 17), name = "") +
  scale_y_log10() +
  labs(title = "Input tissue weight vs. glycogen signal",
       x = "Ctenidia tissue weight (mg)", y = NULL) +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        strip.text = element_text(face = "bold"))

cat("--- weight_check: long-format data for the two-panel weight check ---\n\n")
str(weight_check)
cat("\n--- weight_plot: ggplot object structure ---\n\n")
summary(weight_plot)

ggsave(file.path(output_dir, "tissue_weight_vs_glycogen.png"), weight_plot,
       width = 10, height = 4.5, dpi = 300)

weight_plot

cat("Spearman correlation, tissue weight vs. homogenate glycogen concentration:\n\n")
print(cor.test(sample_glycogen$weight_mg, sample_glycogen$homogenate_conc_ug_mL,
               method = "spearman"))
cat("\nSpearman correlation, tissue weight vs. weight-normalized glycogen:\n\n")
print(cor.test(sample_glycogen$weight_mg, sample_glycogen$norm_glycogen,
               method = "spearman"))
--- weight_check: long-format data for the two-panel weight check ---

tibble [64 × 6] (S3: tbl_df/tbl/data.frame)
 $ sample_id  : chr [1:64] "1_01_ambient" "1_01_ambient" "1_02_ambient" "1_02_ambient" ...
 $ family     : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ temperature: Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 1 1 ...
 $ weight_mg  : num [1:64] 7 7 4.7 4.7 2.5 2.5 10.7 10.7 7.9 7.9 ...
 $ metric     : chr [1:64] "Homogenate glycogen (ug/mL)" "Normalized glycogen (ug/mL/mg tissue)" "Homogenate glycogen (ug/mL)" "Normalized glycogen (ug/mL/mg tissue)" ...
 $ value      : num [1:64] 82.2 11.7 113.8 24.2 54 ...

--- weight_plot: ggplot object structure ---

data: sample_id, family, temperature, weight_mg, metric, value [64x6]
mapping:  x = ~weight_mg, y = ~value
scales:   colour, shape, y, ymin, ymax, yend, yintercept, ymin_final, ymax_final, lower, middle, upper, y0 
faceting:  ~metric 
-----------------------------------
geom_smooth: na.rm = FALSE, orientation = NA, se = FALSE
stat_smooth: na.rm = FALSE, orientation = NA, se = FALSE, method = lm
position_identity 

mapping: colour = ~family, shape = ~temperature 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity 

Spearman correlation, tissue weight vs. homogenate glycogen concentration:


    Spearman's rank correlation rho

data:  sample_glycogen$weight_mg and sample_glycogen$homogenate_conc_ug_mL
S = 1860.7, p-value = 4.114e-05
alternative hypothesis: true rho is not equal to 0
sample estimates:
      rho 
0.6589619 


Spearman correlation, tissue weight vs. weight-normalized glycogen:


    Spearman's rank correlation rho

data:  sample_glycogen$weight_mg and sample_glycogen$norm_glycogen
S = 3839.5, p-value = 0.09966
alternative hypothesis: true rho is not equal to 0
sample estimates:
     rho 
0.296276

Interpretation. Homogenate glycogen concentration (before dividing by weight) is strongly and significantly correlated with input tissue weight (rho = 0.66, p < 0.0001). This is expected, not a sign of a problem: every sample was homogenized and diluted by the same fixed protocol regardless of starting tissue mass, so a heavier piece of ctenidia simply contributes proportionally more total glycogen into the same assay volume, mechanically inflating homogenate-level concentration. It reflects a feature of the protocol (fixed homogenization/dilution volume, variable input mass), not biological variation in glycogen content per se.

Dividing by tissue weight is meant to remove exactly this scaling artifact. After normalization, the correlation with weight drops sharply, to a weak and non-significant residual (rho = 0.30, p = 0.10) – roughly a threefold reduction in the strength of association, and no longer distinguishable from no relationship at this sample size (n = 32). This is the expected signature of normalization working as intended: most, though perhaps not quite all, of the weight-driven scaling has been removed, leaving normalized glycogen approximately independent of how much tissue went into each homogenate. The residual rho = 0.30 is small enough that it does not indicate normalization failure, but it is also not exactly zero, so a mild weight-related trend in the normalized values cannot be fully excluded from this dataset alone (possible contributors include measurement noise at low tissue weights, where small absolute pipetting/weighing errors have larger proportional effects, or a true small biological effect such as larger individuals tending to have either richer or more dilute ctenidia glycogen stores). Because none of the family or temperature comparisons above showed a significant effect, this residual weight association is not currently a confound for the reported group comparisons, but it is worth re-checking if a future dataset shows a significant family or temperature effect that also happens to coincide with a weight imbalance between groups.

6 SAMPLE GLYCOGEN TABLE

sample_table <- sample_glycogen %>%
  transmute(
    Sample        = sample_id,
    Family        = family,
    Temperature   = temperature,
    Individual    = individual,
    `Tissue (mg)` = weight_mg,
    `Dilution`    = dilution,
    `Mean lum.`   = round(mean_lum, 0),
    `CV (%)`      = round(cv_lum, 1),
    `Well glycogen (ug/mL)`        = round(well_conc_ug_mL, 3),
    `Homogenate glycogen (ug/mL)`  = round(homogenate_conc_ug_mL, 2),
    `Normalized glycogen (ug/mL/mg)` = round(norm_glycogen, 2),
    `In std range` = ifelse(in_std_range, "yes", "NO - extrapolated")
  )

kable(sample_table, caption = "Per-sample glycogen quantification (mean of technical triplicate)")

write.csv(sample_table, file.path(output_dir, "sample_glycogen_results.csv"), row.names = FALSE)

cat("\n--- sample_table: formatted per-sample results written to CSV ---\n\n")
str(sample_table)
Sample Family Temperature Individual Tissue (mg) Dilution Mean lum. CV (%) Well glycogen (ug/mL) Homogenate glycogen (ug/mL) Normalized glycogen (ug/mL/mg) In std range
1_01_ambient Family 1 ambient 01 7.0 25 10802 5.0 3.290 82.25 11.75 yes
1_02_ambient Family 1 ambient 02 4.7 25 14632 1.3 4.550 113.76 24.20 yes
1_03_ambient Family 1 ambient 03 2.5 25 7373 3.8 2.162 54.05 21.62 yes
1_04_ambient Family 1 ambient 04 10.7 25 39486 6.4 12.727 318.18 29.74 yes
1_05_ambient Family 1 ambient 05 7.9 25 22099 2.7 7.007 175.17 22.17 yes
1_06_ambient Family 1 ambient 06 11.0 25 97291 6.0 31.745 793.63 72.15 NO - extrapolated
1_07_ambient Family 1 ambient 07 6.1 25 5076 5.4 1.406 35.15 5.76 yes
1_08_ambient Family 1 ambient 08 24.8 25 200961 0.2 65.853 1646.32 66.38 NO - extrapolated
1_01_36C Family 1 36C 01 9.6 25 31844 5.0 10.213 255.32 26.60 yes
1_02_36C Family 1 36C 02 9.2 25 5679 6.6 1.605 40.12 4.36 yes
1_03_36C Family 1 36C 03 3.1 25 4832 9.9 1.326 33.15 10.69 yes
1_04_36C Family 1 36C 04 12.4 25 5075 30.1 1.406 35.15 2.83 yes
1_05_36C Family 1 36C 05 25.7 25 94104 4.9 30.697 767.42 29.86 NO - extrapolated
1_06_36C Family 1 36C 06 13.0 25 56800 1.9 18.424 460.59 35.43 yes
1_07_36C Family 1 36C 07 6.1 25 21660 4.1 6.862 171.56 28.12 yes
1_08_36C Family 1 36C 08 8.8 25 46235 3.5 14.948 373.69 42.46 yes
9_01_ambient Family 9 ambient 01 7.3 25 7336 3.3 2.150 53.75 7.36 yes
9_02_ambient Family 9 ambient 02 4.4 25 4991 4.7 1.378 34.45 7.83 yes
9_03_ambient Family 9 ambient 03 4.3 25 11467 5.1 3.509 87.72 20.40 yes
9_04_ambient Family 9 ambient 04 1.8 25 5653 6.4 1.596 39.90 22.17 yes
9_05_ambient Family 9 ambient 05 9.2 25 70209 1.1 22.835 570.88 62.05 NO - extrapolated
9_06_ambient Family 9 ambient 06 4.3 25 3649 6.5 0.937 23.42 5.45 yes
9_07_ambient Family 9 ambient 07 2.4 25 9969 4.9 3.016 75.40 31.42 yes
9_08_ambient Family 9 ambient 08 7.4 25 2678 5.3 0.617 15.43 2.09 yes
9_01_36C Family 9 36C 01 7.7 25 14413 5.9 4.478 111.95 14.54 yes
9_02_36C Family 9 36C 02 15.7 25 12523 4.4 3.856 96.41 6.14 yes
9_03_36C Family 9 36C 03 10.4 25 43857 3.2 14.165 354.13 34.05 yes
9_04_36C Family 9 36C 04 11.3 25 23983 2.2 7.627 190.67 16.87 yes
9_05_36C Family 9 36C 05 9.2 25 49781 4.3 16.114 402.85 43.79 yes
9_06_36C Family 9 36C 06 4.3 25 14106 3.2 4.377 109.43 25.45 yes
9_07_36C Family 9 36C 07 2.4 25 2408 11.0 0.528 13.21 5.50 yes
9_08_36C Family 9 36C 08 4.9 25 18651 6.6 5.872 146.81 29.96 yes

Per-sample glycogen quantification (mean of technical triplicate)

--- sample_table: formatted per-sample results written to CSV ---

tibble [32 × 12] (S3: tbl_df/tbl/data.frame)
 $ Sample                        : chr [1:32] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ Family                        : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ Temperature                   : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 1 1 2 2 ...
 $ Individual                    : chr [1:32] "01" "02" "03" "04" ...
 $ Tissue (mg)                   : num [1:32] 7 4.7 2.5 10.7 7.9 11 6.1 24.8 9.6 9.2 ...
 $ Dilution                      : num [1:32] 25 25 25 25 25 25 25 25 25 25 ...
 $ Mean lum.                     : num [1:32] 10802 14632 7373 39486 22099 ...
 $ CV (%)                        : num [1:32] 5 1.3 3.8 6.4 2.7 6 5.4 0.2 5 6.6 ...
 $ Well glycogen (ug/mL)         : num [1:32] 3.29 4.55 2.16 12.73 7.01 ...
 $ Homogenate glycogen (ug/mL)   : num [1:32] 82.2 113.8 54 318.2 175.2 ...
 $ Normalized glycogen (ug/mL/mg): num [1:32] 11.8 24.2 21.6 29.7 22.2 ...
 $ In std range                  : chr [1:32] "yes" "yes" "yes" "yes" ...

7 FAMILY AND TEMPERATURE COMPARISONS

7.1 Group summary statistics

group_summary <- sample_glycogen %>%
  group_by(Family = family, Temperature = temperature) %>%
  summarise(
    n      = n(),
    mean   = mean(norm_glycogen),
    sd     = sd(norm_glycogen),
    se     = sd(norm_glycogen) / sqrt(n()),
    median = median(norm_glycogen),
    min    = min(norm_glycogen),
    max    = max(norm_glycogen),
    .groups = "drop"
  )

kable(group_summary, digits = 2,
      caption = "Normalized glycogen (ug/mL per mg tissue) by family and temperature")

write.csv(group_summary, file.path(output_dir, "group_summary_stats.csv"), row.names = FALSE)

cat("\nSame summary excluding the 4 extrapolated samples:\n\n")
group_summary_inrange <- sample_glycogen %>%
  filter(in_std_range) %>%
  group_by(Family = family, Temperature = temperature) %>%
  summarise(n = n(), mean = mean(norm_glycogen), sd = sd(norm_glycogen),
            median = median(norm_glycogen), .groups = "drop")
print(as.data.frame(group_summary_inrange), row.names = FALSE, digits = 4)

cat("\n--- group_summary: family x temperature summary, all 32 samples ---\n\n")
str(group_summary)
cat("\n--- group_summary_inrange: same summary, extrapolated samples excluded ---\n\n")
str(group_summary_inrange)
Family Temperature n mean sd se median min max
Family 1 ambient 8 31.72 24.38 8.62 23.19 5.76 72.15
Family 1 36C 8 22.55 14.75 5.22 27.36 2.83 42.46
Family 9 ambient 8 19.84 19.80 7.00 14.12 2.09 62.05
Family 9 36C 8 22.04 13.63 4.82 21.16 5.50 43.79

Normalized glycogen (ug/mL per mg tissue) by family and temperature

Same summary excluding the 4 extrapolated samples:

   Family Temperature n  mean     sd median
 Family 1     ambient 6 19.21  8.795  21.90
 Family 1         36C 7 21.50 15.612  26.60
 Family 9     ambient 7 13.82 10.863   7.83
 Family 9         36C 8 22.04 13.632  21.16

--- group_summary: family x temperature summary, all 32 samples ---

tibble [4 × 9] (S3: tbl_df/tbl/data.frame)
 $ Family     : Factor w/ 2 levels "Family 1","Family 9": 1 1 2 2
 $ Temperature: Factor w/ 2 levels "ambient","36C": 1 2 1 2
 $ n          : int [1:4] 8 8 8 8
 $ mean       : num [1:4] 31.7 22.5 19.8 22
 $ sd         : num [1:4] 24.4 14.8 19.8 13.6
 $ se         : num [1:4] 8.62 5.22 7 4.82
 $ median     : num [1:4] 23.2 27.4 14.1 21.2
 $ min        : num [1:4] 5.76 2.83 2.09 5.5
 $ max        : num [1:4] 72.1 42.5 62.1 43.8

--- group_summary_inrange: same summary, extrapolated samples excluded ---

tibble [4 × 6] (S3: tbl_df/tbl/data.frame)
 $ Family     : Factor w/ 2 levels "Family 1","Family 9": 1 1 2 2
 $ Temperature: Factor w/ 2 levels "ambient","36C": 1 2 1 2
 $ n          : int [1:4] 6 7 7 8
 $ mean       : num [1:4] 19.2 21.5 13.8 22
 $ sd         : num [1:4] 8.8 15.6 10.9 13.6
 $ median     : num [1:4] 21.9 26.6 7.83 21.16

7.2 Plot: glycogen by family and temperature

group_plot <- ggplot(sample_glycogen,
                     aes(x = temperature, y = norm_glycogen, fill = family)) +
  geom_boxplot(outlier.shape = NA, alpha = 0.35, width = 0.6,
               position = position_dodge(width = 0.75)) +
  geom_point(aes(color = family, shape = in_std_range, group = family),
             position = position_jitterdodge(jitter.width = 0.15, dodge.width = 0.75,
                                             seed = 42),
             size = 2.8) +
  scale_fill_manual(values = c("Family 1" = "#1b7837", "Family 9" = "#762a83"), name = "") +
  scale_color_manual(values = c("Family 1" = "#1b7837", "Family 9" = "#762a83"), name = "") +
  scale_shape_manual(values = c("TRUE" = 16, "FALSE" = 4),
                     labels = c("TRUE" = "within standard curve",
                                "FALSE" = "extrapolated"),
                     name = "") +
  labs(
    title = "Ctenidia glycogen content by USDA family and temperature exposure",
    subtitle = "Each point is one individual (mean of technical triplicate); n = 8 per group",
    x = "Temperature exposure",
    y = "Normalized glycogen (ug/mL per mg tissue)",
    caption = "x = value extrapolated above the 20 ug/mL top standard"
  ) +
  theme_bw() +
  theme(
    plot.title  = element_text(size = 14, face = "bold"),
    axis.title  = element_text(size = 12, face = "bold"),
    legend.position = "bottom"
  )

cat("--- group_plot: ggplot object structure ---\n\n")
summary(group_plot)

ggsave(file.path(output_dir, "glycogen_by_family_temperature.png"), group_plot,
       width = 9, height = 6, dpi = 300)

group_plot

--- group_plot: ggplot object structure ---

data: sample_id, family, individual, temperature, weight_mg, dilution,
  n_reps, mean_lum, sd_lum, cv_lum, well_conc_ug_mL,
  homogenate_conc_ug_mL, norm_glycogen, in_std_range [32x14]
mapping:  x = ~temperature, y = ~norm_glycogen, fill = ~family
scales:   fill, colour, shape 
faceting:  <empty> 
-----------------------------------
geom_boxplot: outliers = TRUE, outlier_gp = list(colour = NULL, fill = NULL, shape = NA, size = NULL, stroke = 0.5, alpha = NULL), whisker_gp = list(colour = NULL, linetype = NULL, linewidth = NULL), staple_gp = list(colour = NULL, linetype = NULL, linewidth = NULL), median_gp = list(colour = NULL, linetype = NULL, linewidth = NULL), box_gp = list(colour = NULL, linetype = NULL, linewidth = NULL), notch = FALSE, notchwidth = 0.5, staplewidth = 0, varwidth = FALSE, na.rm = FALSE, orientation = NA
stat_boxplot: na.rm = FALSE, orientation = NA, width = 0.6
position_dodge 

mapping: colour = ~family, shape = ~in_std_range, group = ~family 
geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_jitterdodge

Note: this boxplot uses a linear y-axis. Given the ~30-fold range of values and the four extrapolated high points, the linear scale visually compresses the lower/mid-range groups toward the bottom of the plot; a log-scale version was reviewed and rejected in favor of this one, but keep the compression in mind when comparing box heights/positions across groups by eye – the statistical tests below are unaffected either way, since they are run on log10-transformed values regardless of which scale is plotted.

7.3 Plot: family x temperature interaction

The boxplot above shows the full per-individual distribution, which makes it hard to visually judge whether the family gap widens, narrows, or reverses with temperature – exactly the question an interaction test asks. The plot below isolates that comparison: group means +/- SEM only, connected by family, so that the two line slopes are a visual proxy for the interaction effect size being tested in the ANOVA (Family x Temperature, p = 0.268 on log10-transformed values; see Statistical tests below). Near-parallel lines indicate no interaction; converging, diverging, or crossing lines would indicate the families respond differently to temperature.

interaction_data <- sample_glycogen %>%
  group_by(family, temperature) %>%
  summarise(mean_glyc = mean(norm_glycogen),
            se_glyc   = sd(norm_glycogen) / sqrt(n()),
            .groups = "drop")

interaction_plot <- ggplot(interaction_data,
                           aes(x = temperature, y = mean_glyc,
                               color = family, group = family)) +
  geom_line(linewidth = 1.1) +
  geom_errorbar(aes(ymin = mean_glyc - se_glyc, ymax = mean_glyc + se_glyc),
                width = 0.08, linewidth = 0.9) +
  geom_point(size = 4) +
  scale_color_manual(values = c("Family 1" = "#1b7837", "Family 9" = "#762a83"), name = "") +
  labs(
    title = "Family x temperature interaction",
    subtitle = "Group means +/- SEM, linear scale (n = 8 per group)",
    x = "Temperature exposure",
    y = "Mean normalized glycogen (ug/mL/mg tissue)"
  ) +
  theme_bw() +
  theme(plot.title = element_text(size = 13, face = "bold"),
        axis.title = element_text(size = 11, face = "bold"))

cat("--- interaction_data: group means and SEM plotted below ---\n")
str(interaction_data)
cat("\n--- interaction_plot: ggplot object structure ---\n")
summary(interaction_plot)

ggsave(file.path(output_dir, "glycogen_family_temperature_interaction.png"), interaction_plot,
       width = 8, height = 5.5, dpi = 300)

interaction_plot

--- interaction_data: group means and SEM plotted below ---
tibble [4 × 4] (S3: tbl_df/tbl/data.frame)
 $ family     : Factor w/ 2 levels "Family 1","Family 9": 1 1 2 2
 $ temperature: Factor w/ 2 levels "ambient","36C": 1 2 1 2
 $ mean_glyc  : num [1:4] 31.7 22.5 19.8 22
 $ se_glyc    : num [1:4] 8.62 5.22 7 4.82

--- interaction_plot: ggplot object structure ---
data: family, temperature, mean_glyc, se_glyc [4x4]
mapping:  x = ~temperature, y = ~mean_glyc, colour = ~family, group = ~family
scales:   colour 
faceting:  <empty> 
-----------------------------------
geom_line: na.rm = FALSE, orientation = NA, arrow = NULL, arrow.fill = NULL, lineend = butt, linejoin = round, linemitre = 10
stat_identity: na.rm = FALSE
position_identity 

mapping: ymin = ~mean_glyc - se_glyc, ymax = ~mean_glyc + se_glyc 
geom_errorbar: na.rm = FALSE, orientation = NA, lineend = butt, width = 0.08
stat_identity: na.rm = FALSE
position_identity 

geom_point: na.rm = FALSE
stat_identity: na.rm = FALSE
position_identity

7.4 Statistical tests

Normalized glycogen spans roughly two orders of magnitude and is strongly right-skewed, so residuals from a two-way ANOVA on the raw scale are non-normal. A log10 transform fixes this, so the log-scale ANOVA is the primary test and the raw-scale ANOVA plus distribution-free rank tests are reported alongside it.

aov_raw <- aov(norm_glycogen ~ family * temperature, data = sample_glycogen)
aov_log <- aov(log10(norm_glycogen) ~ family * temperature, data = sample_glycogen)

cat("Shapiro-Wilk on ANOVA residuals:\n")
cat("  raw scale:   p =", signif(shapiro.test(residuals(aov_raw))$p.value, 4), "\n")
cat("  log10 scale: p =", signif(shapiro.test(residuals(aov_log))$p.value, 4), "\n")
cat("\nBartlett test of homogeneity of variance (log10 scale): p =",
    signif(bartlett.test(log10(norm_glycogen) ~ interaction(family, temperature),
                         data = sample_glycogen)$p.value, 4), "\n")

cat("\n--- aov_raw: two-way ANOVA model, raw scale ---\n")
str(aov_raw, max.level = 1, give.attr = FALSE)
cat("\n--- aov_log: two-way ANOVA model, log10 scale (primary test) ---\n")
str(aov_log, max.level = 1, give.attr = FALSE)
Shapiro-Wilk on ANOVA residuals:
  raw scale:   p = 0.02786 
  log10 scale: p = 0.1979 

Bartlett test of homogeneity of variance (log10 scale): p = 0.7831 

--- aov_raw: two-way ANOVA model, raw scale ---
List of 13
 $ coefficients : Named num [1:4] 31.72 -11.88 -9.18 11.37
 $ residuals    : Named num [1:32] -19.97 -7.52 -10.1 -1.99 -9.55 ...
 $ effects      : Named num [1:32] -135.98 -17.51 -9.88 -16.08 -5.45 ...
 $ rank         : int 4
 $ fitted.values: Named num [1:32] 31.7 31.7 31.7 31.7 31.7 ...
 $ assign       : int [1:4] 0 1 2 3
 $ qr           :List of 5
 $ df.residual  : int 28
 $ contrasts    :List of 2
 $ xlevels      :List of 2
 $ call         : language aov(formula = norm_glycogen ~ family * temperature, data = sample_glycogen)
 $ terms        :Classes 'terms', 'formula'  language norm_glycogen ~ family * temperature
 $ model        :'data.frame':  32 obs. of  3 variables:

--- aov_log: two-way ANOVA model, log10 scale (primary test) ---
List of 13
 $ coefficients : Named num [1:4] 1.381 -0.286 -0.175 0.327
 $ residuals    : Named num [1:32] -0.31106 0.00278 -0.04627 0.09219 -0.03526 ...
 $ effects      : Named num [1:32] -6.97065 -0.3466 -0.03325 -0.46218 -0.00809 ...
 $ rank         : int 4
 $ fitted.values: Named num [1:32] 1.38 1.38 1.38 1.38 1.38 ...
 $ assign       : int [1:4] 0 1 2 3
 $ qr           :List of 5
 $ df.residual  : int 28
 $ contrasts    :List of 2
 $ xlevels      :List of 2
 $ call         : language aov(formula = log10(norm_glycogen) ~ family * temperature, data = sample_glycogen)
 $ terms        :Classes 'terms', 'formula'  language log10(norm_glycogen) ~ family * temperature
 $ model        :'data.frame':  32 obs. of  3 variables:
cat("=== Two-way ANOVA on log10(normalized glycogen) [primary] ===\n")
print(summary(aov_log))

cat("\n=== Two-way ANOVA on raw normalized glycogen [for reference] ===\n")
print(summary(aov_raw))
=== Two-way ANOVA on log10(normalized glycogen) [primary] ===
                   Df Sum Sq Mean Sq F value Pr(>F)
family              1  0.120 0.12013   0.719  0.404
temperature         1  0.001 0.00111   0.007  0.936
family:temperature  1  0.214 0.21361   1.278  0.268
Residuals          28  4.681 0.16717               

=== Two-way ANOVA on raw normalized glycogen [for reference] ===
                   Df Sum Sq Mean Sq F value Pr(>F)
family              1    307   306.7   0.883  0.356
temperature         1     98    97.5   0.281  0.600
family:temperature  1    259   258.6   0.744  0.396
Residuals          28   9730   347.5               
cat("=== Wilcoxon rank-sum: family 1 vs family 9 (temperatures pooled) ===\n")
print(wilcox.test(norm_glycogen ~ family, data = sample_glycogen))

cat("=== Wilcoxon rank-sum: ambient vs 36C (families pooled) ===\n")
print(wilcox.test(norm_glycogen ~ temperature, data = sample_glycogen))

cat("=== Stratified comparisons ===\n")
strat <- rbind(
  data.frame(comparison = "ambient vs 36C, within family 1",
             p = wilcox.test(norm_glycogen ~ temperature,
                             data = filter(sample_glycogen, family == "Family 1"))$p.value),
  data.frame(comparison = "ambient vs 36C, within family 9",
             p = wilcox.test(norm_glycogen ~ temperature,
                             data = filter(sample_glycogen, family == "Family 9"))$p.value),
  data.frame(comparison = "family 1 vs 9, within ambient",
             p = wilcox.test(norm_glycogen ~ family,
                             data = filter(sample_glycogen, temperature == "ambient"))$p.value),
  data.frame(comparison = "family 1 vs 9, within 36C",
             p = wilcox.test(norm_glycogen ~ family,
                             data = filter(sample_glycogen, temperature == "36C"))$p.value)
)
strat$p_holm <- p.adjust(strat$p, method = "holm")
kable(strat, digits = 4,
      col.names = c("Comparison", "p (Wilcoxon)", "p (Holm-adjusted)"),
      caption = "Stratified pairwise comparisons of normalized glycogen")

cat("\n--- strat: stratified Wilcoxon comparisons with Holm-adjusted p-values ---\n")
str(strat)
=== Wilcoxon rank-sum: family 1 vs family 9 (temperatures pooled) ===

    Wilcoxon rank sum exact test

data:  norm_glycogen by family
W = 150, p-value = 0.423
alternative hypothesis: true location shift is not equal to 0

=== Wilcoxon rank-sum: ambient vs 36C (families pooled) ===

    Wilcoxon rank sum exact test

data:  norm_glycogen by temperature
W = 123, p-value = 0.8672
alternative hypothesis: true location shift is not equal to 0

=== Stratified comparisons ===
Comparison p (Wilcoxon) p (Holm-adjusted)
ambient vs 36C, within family 1 0.7984 1.0000
ambient vs 36C, within family 9 0.5737 1.0000
family 1 vs 9, within ambient 0.1949 0.7795
family 1 vs 9, within 36C 0.9591 1.0000

Stratified pairwise comparisons of normalized glycogen

--- strat: stratified Wilcoxon comparisons with Holm-adjusted p-values ---
'data.frame':   4 obs. of  3 variables:
 $ comparison: chr  "ambient vs 36C, within family 1" "ambient vs 36C, within family 9" "family 1 vs 9, within ambient" "family 1 vs 9, within 36C"
 $ p         : num  0.798 0.574 0.195 0.959
 $ p_holm    : num  1 1 0.779 1

7.4.1 Sensitivity analysis: excluding extrapolated samples

Three of the four samples that read above the top standard are family 1 (1_08_ambient, 1_06_ambient, 1_05_36C) and one is family 9 (9_05_ambient), so it is important to confirm they are not driving any apparent family or temperature effect.

in_range_only <- filter(sample_glycogen, in_std_range)

cat("n retained:", nrow(in_range_only), "of", nrow(sample_glycogen), "\n\n")
cat("=== Two-way ANOVA on log10(normalized glycogen), in-range samples only ===\n")
print(summary(aov(log10(norm_glycogen) ~ family * temperature, data = in_range_only)))

cat("\n=== Wilcoxon, family 1 vs 9 (in-range only, temperatures pooled) ===\n")
print(wilcox.test(norm_glycogen ~ family, data = in_range_only))

cat("\n--- in_range_only: sample_glycogen with extrapolated samples removed ---\n")
str(in_range_only)
n retained: 28 of 32 

=== Two-way ANOVA on log10(normalized glycogen), in-range samples only ===
                   Df Sum Sq Mean Sq F value Pr(>F)
family              1  0.030 0.03027   0.209  0.652
temperature         1  0.079 0.07915   0.546  0.467
family:temperature  1  0.168 0.16849   1.162  0.292
Residuals          24  3.479 0.14497               

=== Wilcoxon, family 1 vs 9 (in-range only, temperatures pooled) ===

    Wilcoxon rank sum exact test

data:  norm_glycogen by family
W = 107, p-value = 0.6832
alternative hypothesis: true location shift is not equal to 0


--- in_range_only: sample_glycogen with extrapolated samples removed ---
tibble [28 × 14] (S3: tbl_df/tbl/data.frame)
 $ sample_id            : chr [1:28] "1_01_ambient" "1_02_ambient" "1_03_ambient" "1_04_ambient" ...
 $ family               : Factor w/ 2 levels "Family 1","Family 9": 1 1 1 1 1 1 1 1 1 1 ...
 $ individual           : chr [1:28] "01" "02" "03" "04" ...
 $ temperature          : Factor w/ 2 levels "ambient","36C": 1 1 1 1 1 1 2 2 2 2 ...
 $ weight_mg            : num [1:28] 7 4.7 2.5 10.7 7.9 6.1 9.6 9.2 3.1 12.4 ...
 $ dilution             : num [1:28] 25 25 25 25 25 25 25 25 25 25 ...
 $ n_reps               : int [1:28] 3 3 3 3 3 3 3 3 3 3 ...
 $ mean_lum             : num [1:28] 10802 14632 7373 39486 22099 ...
 $ sd_lum               : num [1:28] 543 189 278 2525 603 ...
 $ cv_lum               : num [1:28] 5.03 1.29 3.77 6.4 2.73 ...
 $ well_conc_ug_mL      : num [1:28] 3.29 4.55 2.16 12.73 7.01 ...
 $ homogenate_conc_ug_mL: num [1:28] 82.2 113.8 54 318.2 175.2 ...
 $ norm_glycogen        : num [1:28] 11.7 24.2 21.6 29.7 22.2 ...
 $ in_std_range         : logi [1:28] TRUE TRUE TRUE TRUE TRUE TRUE ...
# Persist the tidy well-level data alongside the summaries for downstream reuse
write.csv(samples_wells %>%
            select(plate, well, sample_id, family, temperature, individual,
                   weight_mg, dilution, luminescence),
          file.path(output_dir, "well_level_luminescence.csv"), row.names = FALSE)

write.csv(glycogen_summary_data,
          file.path(output_dir, "standard_curve_summary.csv"), row.names = FALSE)

cat("Files written to", output_dir, ":\n")
cat(paste(" -", list.files(output_dir)), sep = "\n")
Files written to ../output/Gen5-20260730-mgig-fams1_9-glycogen-glo :
 - glycogen_by_family_temperature.png
 - glycogen_family_temperature_interaction.png
 - glycogen_standard_curve_loglog.png
 - glycogen_standard_curve.png
 - group_summary_stats.csv
 - plate01_luminescence_map.png
 - sample_glycogen_results.csv
 - standard_curve_summary.csv
 - tissue_weight_vs_glycogen.png
 - well_level_luminescence.csv

8 SUMMARY

Standard curve. The four-point glycogen curve plus zero standard is linear across the full range when fit to concentration means (R² = 0.9999), but the fit is dominated by the single 20 µg/mL point; fit to individual replicate wells R² drops to 0.986, and the 20 µg/mL standard itself has a 12.6% CV. The three lowest standards (0.02-2 µg/mL) are barely resolved above the buffer-only background (~910 RLU). Adding intermediate standards (e.g. 5 and 10 µg/mL) and tightening replicate pipetting at the top standard would make back-calculation in the mid-range considerably more trustworthy.

Technical quality. Triplicate CVs are good overall (median 4.9%, mean 5.5%); only 1_04_36C exceeds 15% (30.1%, driven by one low replicate). The plate map shows no obvious row-wise gradient across the family/temperature row-blocks (A/B, C/D, E/F, G/H), so the row-block design does not appear to be confounded with plate position.

Range. Four samples read above the top standard and are extrapolated: 1_08_ambient, 1_06_ambient, 1_05_36C, 9_05_ambient. 1_08_ambient in particular (~201,000 RLU, 3.3× the top standard) should be re-run at a higher dilution; the other three are more modestly above range (~94,000-97,000 RLU). The extrapolated samples are not concentrated in a single family.

Recommended re-assay dilutions for these four samples, targeting a well-level concentration near the middle of the quantifiable curve (~10 µg/mL, i.e. roughly the midpoint between the 2 and 20 µg/mL standards, corresponding to ~10 µg/mL/mg tissue-normalized once re-scaled by weight) rather than just clearing the 20 µg/mL ceiling:

Sample Mean luminescence (RLU, current 1:25) Current dilution Recommended total dilution Approx. fold increase
1_08_ambient ~201,000 1:25 ~1:150 ~6×
1_06_ambient ~97,300 1:25 ~1:80 ~3×
1_05_36C ~94,100 1:25 ~1:75 ~3×
9_05_ambient ~70,200 1:25 ~1:60 ~2×

These are rough targets based on each sample’s current homogenate-level glycogen estimate (extrapolated off the standard curve) and are intended as a starting point for the next dilution series, not a precise prescription – actual optimal dilution should be confirmed empirically, since the extrapolated homogenate concentrations that these estimates are based on are themselves the values in question.

Family effect. Pooling temperatures, group means are 27.1 (family 1) vs 20.9 (family 9) µg/mL/mg tissue. The effect is not significant in the log-scale two-way ANOVA (F = 0.72, p = 0.404) or by Wilcoxon (W = 150, p = 0.423). None of the four stratified family/temperature contrasts reach significance either before or after Holm correction (smallest unadjusted p = 0.195, within ambient).

Temperature effect. No detectable effect of 36°C exposure. Neither the main effect (log-scale ANOVA p = 0.936; Wilcoxon p = 0.867) nor the family × temperature interaction (p = 0.268) reaches significance. Group means by family and temperature are: family 1 ambient 31.7, family 1 36°C 22.5, family 9 ambient 19.8, family 9 36°C 22.0 µg/mL/mg tissue – no consistent within-family shift with temperature, and no interaction pattern suggestive of family-specific heat response.

Weight relationship. Homogenate glycogen concentration correlates with input tissue weight (Spearman rho = 0.66, p < 0.0001), as expected since larger samples were run at the same fixed dilution rather than volume-adjusted – this is a protocol artifact of fixed homogenization volume, not a biological signal. Weight-normalized glycogen shows a much weaker, non-significant residual correlation with weight (rho = 0.30, p = 0.10), consistent with the normalization step removing most, though perhaps not all, of that scaling effect. The residual association is small enough not to indicate a normalization failure, and since neither family nor temperature showed a significant effect in this dataset, it is not currently confounding those comparisons – but it would be worth revisiting if a future run shows a significant group difference alongside a weight imbalance between groups.

Caveats.

  1. None of the family, temperature, or family × temperature effects are statistically significant at n = 8 per group – excluding the four extrapolated samples changes this picture only slightly (log-scale ANOVA family p = 0.652, temperature p = 0.467, interaction p = 0.292, n = 28), so the qualitative conclusion (no detectable family or temperature effect) is robust to that exclusion.
  2. Standards were on a separate plate from the samples, so plate-to-plate signal offset is uncorrected. Including a short standard curve on each sample plate would remove this assumption.
  3. Homogenization volume is not recorded in these files; normalized values are therefore per mL of homogenate per mg tissue. Group comparisons are unaffected provided the volume was constant, but absolute glycogen per mg tissue cannot be stated without it.
]]> 2026 glycogen Glycogen-Glo SORMI Pacific oyster Magallana gigas Crassostrea gigas https://robertslab.github.io/sams-notebook/posts/2026/2026-07-30-Glycogen-Assay---SORMI-June-2026-M.gigas-USDA-Families-1-and-9-Comparisons-Using-Glycogen-Glo-Kit/ Thu, 30 Jul 2026 07:00:00 GMT Homogenization - June 2026 SORMI M.gigas Ctenidia from Families 1 and 9 for Glycogen Glo Assay Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-07-22-Homogenization---June-2026-SORMI-M.gigas-Ctenidia-from-Families-1-and-9-for-Glycogen-Glo-Assay/

INTRO

A large experiment and sampling event was conducted on 20260626 (GitHub repo) as part of the sormi-assay-development project (GitHub repo). As part of this, we decided to evaluate glycogen content in ctenidia from two families of oysters (USDA Families 1 and 9) from the ambient and 36°C treatments using the Glycogen Glo Assay. Family selection was decided by Steven.

MATERIALS & METHODS

Frozen ctenidia samples (stored at -80°C) from Families 1 and 9, ambient and 36°C treatments (n = 8 from each treatment in each family), were placed on dry ice. A small portion of each ctenidia sample was weighed (to the nearest 0.01 g) and placed into a 1.5mL Safe-Lock tube (Eppendorf) containting ~300uL of 0.5mm glass beads (Research Products International) and 500uL of PBS. Immediately after adding tissue, 250uL of 0.3N HCl was added and vortexed, per the Glycogen Glo Assay protocol. The samples were then homogenized in a Bullet Blender 5E Gold+ using the 1.5mL adapters. Homogenization was performed for 10 minutes at Speed 12. After homogenization, 250uL of TRIS Buffer was added to each tube and vortexed, per the Glycogen Glo Assay protocol. Samples were stored at -20°C until further processing.

Table of oyster IDs and weights (mg) of ctenidia samples used for the Glycogen Glo Assay:

oyster_ID weight.mg
1-01-Ambient 7
1-02-Ambient 4.7
1-03-Ambient 2.5
1-04-Ambient 10.7
1-05-Ambient 7.9
1-06-Ambient 11
1-07-Ambient 6.1
1-08-Ambient 24.8
1-01-36C 9.6
1-02-36C 9.2
1-03-36C 3.1
1-04-36C 12.4
1-05-36C 25.7
1-06-36C 13
1-07-36C 6.1
1-08-36C 8.8
9-01-Ambient 7.3
9-02-Ambient 4.4
9-03-Ambient 4.3
9-04-Ambient 1.8
9-05-Ambient 18.6
9-06-Ambient 2.8
9-07-Ambient 5.8
9-08-Ambient 7.4
9-01-36C 7.7
9-02-36C 15.7
9-03-36C 10.4
9-04-36C 11.3
9-05-36C 9.2
9-06-36C 4.3
9-07-36C 2.4
9-08-36C 4.9

For reference, here’s a link to the experiment metadata for the June 2026 SORMI experiment:

]]> 2026 SORMI Magallana gigas ctenidia Pacific oyster Crassostrea gigas glycogen Glycogen Glo homogenization https://robertslab.github.io/sams-notebook/posts/2026/2026-07-22-Homogenization---June-2026-SORMI-M.gigas-Ctenidia-from-Families-1-and-9-for-Glycogen-Glo-Assay/ Wed, 22 Jul 2026 07:00:00 GMT Data Received - Full RNA-seq Data for Andy Dittman and NOAA Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-07-15-Data-Received---Full-RNA-seq-Data-for-Andy-Dittman-and-NOAA/

INTRO

Received RNA-seq data for Andy Dittman at NOAA from UW’s Northwest Genomics Center. This is part of a project initiated a while ago (GitHub Issue). The initial RNA submitted was potentially questionable, based on Bionalyzer RIN scores and a subset of the data was previously received on 20260508 (Notebook entry) and QC’d on 20260519 (Notebook entry). This is the remainder of the data that was sequenced and received.

DATA

Data was downloaded to our server, Owl, via GlobusConnect. It took a full 24hrs to complete.

Data is currently available at:

MD5 checksums were verified:

lib_213507_sample_1703710_S21_L001_R1_001.fastq.gz: OK
lib_213524_sample_1703765_S143_L002_R2_001.fastq.gz: OK
lib_213501_sample_1703771_S148_L002_R2_001.fastq.gz: OK
lib_213492_sample_1703673_S126_L002_R2_001.fastq.gz: OK
lib_213476_sample_1703761_S139_L002_R2_001.fastq.gz: OK
lib_213518_sample_1703702_S14_L001_R2_001.fastq.gz: OK
lib_213562_sample_1703786_S64_L001_R1_001.fastq.gz: OK
lib_213525_sample_1703773_S53_L001_R1_001.fastq.gz: OK
lib_213489_sample_1703770_S147_L002_R1_001.fastq.gz: OK
lib_213557_sample_1703740_S42_L001_R2_001.fastq.gz: OK
lib_213487_sample_1703753_S132_L002_R1_001.fastq.gz: OK
lib_213567_sample_1703719_S26_L001_R1_001.fastq.gz: OK
lib_213554_sample_1703705_S17_L001_R2_001.fastq.gz: OK
lib_213566_sample_1703706_S18_L001_R1_001.fastq.gz: OK
lib_213528_sample_1703679_S129_L002_R2_001.fastq.gz: OK
lib_213513_sample_1703772_S52_L001_R1_001.fastq.gz: OK
lib_213508_sample_1703727_S30_L001_R2_001.fastq.gz: OK
lib_213546_sample_1703749_S49_L001_R2_001.fastq.gz: OK
lib_213485_sample_1703734_S36_L001_R2_001.fastq.gz: OK
lib_213529_sample_1703694_S7_L001_R2_001.fastq.gz: OK
lib_213510_sample_1703744_S46_L001_R1_001.fastq.gz: OK
lib_213565_sample_1703698_S10_L001_R2_001.fastq.gz: OK
lib_213498_sample_1703743_S45_L001_R2_001.fastq.gz: OK
lib_213542_sample_1703704_S16_L001_R1_001.fastq.gz: OK
lib_213508_sample_1703727_S30_L001_R1_001.fastq.gz: OK
lib_213553_sample_1703696_S9_L001_R2_001.fastq.gz: OK
lib_213493_sample_1703689_S4_L001_R2_001.fastq.gz: OK
lib_213550_sample_1703785_S63_L001_R1_001.fastq.gz: OK
lib_213515_sample_1703790_S66_L001_R1_001.fastq.gz: OK
lib_213510_sample_1703744_S46_L001_R2_001.fastq.gz: OK
lib_213504_sample_1703675_S127_L002_R2_001.fastq.gz: OK
lib_213473_sample_1703732_S35_L001_R2_001.fastq.gz: OK
lib_213520_sample_1703728_S31_L001_R1_001.fastq.gz: OK
lib_213517_sample_1703693_S6_L001_R1_001.fastq.gz: OK
lib_213483_sample_1703707_S19_L001_R1_001.fastq.gz: OK
lib_213534_sample_1703747_S48_L001_R2_001.fastq.gz: OK
lib_213541_sample_1703695_S8_L001_R1_001.fastq.gz: OK
lib_213486_sample_1703742_S44_L001_R2_001.fastq.gz: OK
lib_213524_sample_1703765_S143_L002_R1_001.fastq.gz: OK
lib_213545_sample_1703739_S41_L001_R1_001.fastq.gz: OK
lib_213479_sample_1703787_S149_L002_R1_001.fastq.gz: OK
lib_213543_sample_1703715_S24_L001_R1_001.fastq.gz: OK
lib_213496_sample_1703726_S29_L001_R1_001.fastq.gz: OK
lib_213553_sample_1703696_S9_L001_R1_001.fastq.gz: OK
lib_213566_sample_1703706_S18_L001_R2_001.fastq.gz: OK
lib_213476_sample_1703761_S139_L002_R1_001.fastq.gz: OK
lib_213494_sample_1703700_S12_L001_R1_001.fastq.gz: OK
lib_213549_sample_1703775_S55_L001_R2_001.fastq.gz: OK
lib_213539_sample_1703792_S68_L001_R2_001.fastq.gz: OK
lib_213555_sample_1703717_S25_L001_R2_001.fastq.gz: OK
lib_213538_sample_1703784_S62_L001_R1_001.fastq.gz: OK
lib_213498_sample_1703743_S45_L001_R1_001.fastq.gz: OK
lib_213489_sample_1703770_S147_L002_R2_001.fastq.gz: OK
lib_213478_sample_1703777_S57_L001_R1_001.fastq.gz: OK
lib_213497_sample_1703735_S37_L001_R1_001.fastq.gz: OK
lib_213504_sample_1703675_S127_L002_R1_001.fastq.gz: OK
lib_213519_sample_1703711_S22_L001_R2_001.fastq.gz: OK
lib_213537_sample_1703774_S54_L001_R2_001.fastq.gz: OK
lib_213555_sample_1703717_S25_L001_R1_001.fastq.gz: OK
lib_213531_sample_1703713_S23_L001_R2_001.fastq.gz: OK
lib_213522_sample_1703746_S47_L001_R1_001.fastq.gz: OK
lib_213474_sample_1703741_S43_L001_R2_001.fastq.gz: OK
lib_213544_sample_1703730_S33_L001_R2_001.fastq.gz: OK
lib_213532_sample_1703729_S32_L001_R2_001.fastq.gz: OK
lib_213474_sample_1703741_S43_L001_R1_001.fastq.gz: OK
lib_213543_sample_1703715_S24_L001_R2_001.fastq.gz: OK
lib_213514_sample_1703781_S60_L001_R2_001.fastq.gz: OK
lib_213497_sample_1703735_S37_L001_R2_001.fastq.gz: OK
lib_213516_sample_1703676_S128_L002_R2_001.fastq.gz: OK
lib_213509_sample_1703736_S38_L001_R2_001.fastq.gz: OK
lib_213558_sample_1703750_S50_L001_R1_001.fastq.gz: OK
lib_213490_sample_1703778_S58_L001_R2_001.fastq.gz: OK
lib_213517_sample_1703693_S6_L001_R2_001.fastq.gz: OK
lib_213522_sample_1703746_S47_L001_R2_001.fastq.gz: OK
lib_213487_sample_1703753_S132_L002_R2_001.fastq.gz: OK
lib_213539_sample_1703792_S68_L001_R1_001.fastq.gz: OK
lib_213530_sample_1703703_S15_L001_R1_001.fastq.gz: OK
lib_213514_sample_1703781_S60_L001_R1_001.fastq.gz: OK
lib_213495_sample_1703709_S20_L001_R2_001.fastq.gz: OK
lib_213482_sample_1703699_S11_L001_R1_001.fastq.gz: OK
lib_213473_sample_1703732_S35_L001_R1_001.fastq.gz: OK
lib_213494_sample_1703700_S12_L001_R2_001.fastq.gz: OK
lib_213513_sample_1703772_S52_L001_R2_001.fastq.gz: OK
lib_213552_sample_1703684_S1_L001_R1_001.fastq.gz: OK
lib_213519_sample_1703711_S22_L001_R1_001.fastq.gz: OK
lib_213503_sample_1703789_S65_L001_R2_001.fastq.gz: OK
lib_213565_sample_1703698_S10_L001_R1_001.fastq.gz: OK
lib_213546_sample_1703749_S49_L001_R1_001.fastq.gz: OK
lib_213525_sample_1703773_S53_L001_R2_001.fastq.gz: OK
lib_213496_sample_1703726_S29_L001_R2_001.fastq.gz: OK
lib_213560_sample_1703768_S146_L002_R1_001.fastq.gz: OK
lib_213521_sample_1703737_S39_L001_R1_001.fastq.gz: OK
lib_213532_sample_1703729_S32_L001_R1_001.fastq.gz: OK
lib_213482_sample_1703699_S11_L001_R2_001.fastq.gz: OK
lib_213521_sample_1703737_S39_L001_R2_001.fastq.gz: OK
lib_213533_sample_1703738_S40_L001_R1_001.fastq.gz: OK
lib_213551_sample_1703793_S69_L001_R2_001.fastq.gz: OK
lib_213516_sample_1703676_S128_L002_R1_001.fastq.gz: OK
lib_213545_sample_1703739_S41_L001_R2_001.fastq.gz: OK
lib_213540_sample_1703680_S130_L002_R2_001.fastq.gz: OK
lib_213560_sample_1703768_S146_L002_R2_001.fastq.gz: OK
lib_213564_sample_1703685_S2_L001_R1_001.fastq.gz: OK
lib_213548_sample_1703767_S145_L002_R2_001.fastq.gz: OK
lib_213523_sample_1703756_S135_L002_R2_001.fastq.gz: OK
lib_213475_sample_1703751_S131_L002_R1_001.fastq.gz: OK
lib_213533_sample_1703738_S40_L001_R2_001.fastq.gz: OK
lib_213531_sample_1703713_S23_L001_R1_001.fastq.gz: OK
lib_213512_sample_1703764_S142_L002_R1_001.fastq.gz: OK
lib_213536_sample_1703766_S144_L002_R2_001.fastq.gz: OK
lib_213528_sample_1703679_S129_L002_R1_001.fastq.gz: OK
lib_213559_sample_1703760_S138_L002_R2_001.fastq.gz: OK
lib_213534_sample_1703747_S48_L001_R1_001.fastq.gz: OK
lib_213535_sample_1703757_S136_L002_R2_001.fastq.gz: OK
lib_213495_sample_1703709_S20_L001_R1_001.fastq.gz: OK
lib_213526_sample_1703782_S61_L001_R1_001.fastq.gz: OK
lib_213550_sample_1703785_S63_L001_R2_001.fastq.gz: OK
lib_213558_sample_1703750_S50_L001_R2_001.fastq.gz: OK
lib_213523_sample_1703756_S135_L002_R1_001.fastq.gz: OK
lib_213511_sample_1703755_S134_L002_R2_001.fastq.gz: OK
lib_213556_sample_1703731_S34_L001_R1_001.fastq.gz: OK
lib_213491_sample_1703788_S150_L002_R1_001.fastq.gz: OK
lib_213559_sample_1703760_S138_L002_R1_001.fastq.gz: OK
lib_213483_sample_1703707_S19_L001_R2_001.fastq.gz: OK
lib_213551_sample_1703793_S69_L001_R1_001.fastq.gz: OK
lib_213505_sample_1703692_S5_L001_R2_001.fastq.gz: OK
lib_213544_sample_1703730_S33_L001_R1_001.fastq.gz: OK
lib_213547_sample_1703758_S137_L002_R2_001.fastq.gz: OK
lib_213529_sample_1703694_S7_L001_R1_001.fastq.gz: OK
lib_213503_sample_1703789_S65_L001_R1_001.fastq.gz: OK
lib_213500_sample_1703763_S141_L002_R1_001.fastq.gz: OK
lib_213515_sample_1703790_S66_L001_R2_001.fastq.gz: OK
lib_213499_sample_1703754_S133_L002_R2_001.fastq.gz: OK
lib_213556_sample_1703731_S34_L001_R2_001.fastq.gz: OK
lib_213492_sample_1703673_S126_L002_R1_001.fastq.gz: OK
lib_213506_sample_1703701_S13_L001_R1_001.fastq.gz: OK
lib_213567_sample_1703719_S26_L001_R2_001.fastq.gz: OK
lib_213472_sample_1703724_S27_L001_R1_001.fastq.gz: OK
lib_213518_sample_1703702_S14_L001_R1_001.fastq.gz: OK
lib_213485_sample_1703734_S36_L001_R1_001.fastq.gz: OK
lib_213493_sample_1703689_S4_L001_R1_001.fastq.gz: OK
lib_213554_sample_1703705_S17_L001_R1_001.fastq.gz: OK
lib_213478_sample_1703777_S57_L001_R2_001.fastq.gz: OK
lib_213537_sample_1703774_S54_L001_R1_001.fastq.gz: OK
lib_213472_sample_1703724_S27_L001_R2_001.fastq.gz: OK
lib_213507_sample_1703710_S21_L001_R2_001.fastq.gz: OK
lib_213502_sample_1703779_S59_L001_R1_001.fastq.gz: OK
lib_213502_sample_1703779_S59_L001_R2_001.fastq.gz: OK
lib_213535_sample_1703757_S136_L002_R1_001.fastq.gz: OK
lib_213501_sample_1703771_S148_L002_R1_001.fastq.gz: OK
lib_213490_sample_1703778_S58_L001_R1_001.fastq.gz: OK
lib_213547_sample_1703758_S137_L002_R1_001.fastq.gz: OK
lib_213509_sample_1703736_S38_L001_R1_001.fastq.gz: OK
lib_213484_sample_1703725_S28_L001_R2_001.fastq.gz: OK
lib_213491_sample_1703788_S150_L002_R2_001.fastq.gz: OK
lib_213486_sample_1703742_S44_L001_R1_001.fastq.gz: OK
lib_213499_sample_1703754_S133_L002_R1_001.fastq.gz: OK
lib_213527_sample_1703791_S67_L001_R1_001.fastq.gz: OK
lib_213475_sample_1703751_S131_L002_R2_001.fastq.gz: OK
lib_213506_sample_1703701_S13_L001_R2_001.fastq.gz: OK
lib_213480_sample_1703672_S125_L002_R1_001.fastq.gz: OK
lib_213538_sample_1703784_S62_L001_R2_001.fastq.gz: OK
lib_213481_sample_1703686_S3_L001_R2_001.fastq.gz: OK
lib_213477_sample_1703769_S51_L001_R1_001.fastq.gz: OK
lib_213526_sample_1703782_S61_L001_R2_001.fastq.gz: OK
lib_213564_sample_1703685_S2_L001_R2_001.fastq.gz: OK
lib_213563_sample_1703794_S70_L001_R1_001.fastq.gz: OK
lib_213477_sample_1703769_S51_L001_R2_001.fastq.gz: OK
lib_213549_sample_1703775_S55_L001_R1_001.fastq.gz: OK
lib_213536_sample_1703766_S144_L002_R1_001.fastq.gz: OK
lib_213511_sample_1703755_S134_L002_R1_001.fastq.gz: OK
lib_213540_sample_1703680_S130_L002_R1_001.fastq.gz: OK
lib_213480_sample_1703672_S125_L002_R2_001.fastq.gz: OK
lib_213520_sample_1703728_S31_L001_R2_001.fastq.gz: OK
lib_213505_sample_1703692_S5_L001_R1_001.fastq.gz: OK
lib_213488_sample_1703762_S140_L002_R1_001.fastq.gz: OK
lib_213552_sample_1703684_S1_L001_R2_001.fastq.gz: OK
lib_213512_sample_1703764_S142_L002_R2_001.fastq.gz: OK
lib_213488_sample_1703762_S140_L002_R2_001.fastq.gz: OK
lib_213479_sample_1703787_S149_L002_R2_001.fastq.gz: OK
lib_213561_sample_1703776_S56_L001_R1_001.fastq.gz: OK
lib_213541_sample_1703695_S8_L001_R2_001.fastq.gz: OK
lib_213548_sample_1703767_S145_L002_R1_001.fastq.gz: OK
lib_213557_sample_1703740_S42_L001_R1_001.fastq.gz: OK
lib_213542_sample_1703704_S16_L001_R2_001.fastq.gz: OK
lib_213561_sample_1703776_S56_L001_R2_001.fastq.gz: OK
lib_213500_sample_1703763_S141_L002_R2_001.fastq.gz: OK
lib_213530_sample_1703703_S15_L001_R2_001.fastq.gz: OK
lib_213562_sample_1703786_S64_L001_R2_001.fastq.gz: OK
lib_213563_sample_1703794_S70_L001_R2_001.fastq.gz: OK
lib_213527_sample_1703791_S67_L001_R2_001.fastq.gz: OK
lib_213484_sample_1703725_S28_L001_R1_001.fastq.gz: OK
lib_213481_sample_1703686_S3_L001_R1_001.fastq.gz: OK
]]> Data Received RNA-seq 2026 https://robertslab.github.io/sams-notebook/posts/2026/2026-07-15-Data-Received---Full-RNA-seq-Data-for-Andy-Dittman-and-NOAA/ Wed, 15 Jul 2026 07:00:00 GMT Resazurin Assays - USDA M.gigas Families in Response to Temperature Stress Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-07-13-Resazurin-Assays---USDA-M.gigas-Families-in-Response-to-Temperature-Stress/

INTRO

Twenty adult oysters from each of four different USDA families (1, 5, 7, & 9) were subjected to acute heat stress at 36oC for four hours in an upright incubator. Oysters were photographed for measurement and then submerged in ~70mL of freshly prepared resazuring solution in plastic drink cups. Resazurin fluorescence was measured by extracting 200uL from each cup and placing in a 96-well plate, then measuring fluorescence in a Synergy HTX plate reader (Agilent).

Image of undergrad loading tray of cups containing oysters and resazurin into incubator in Graham Young's lab

Image of undergrad loading tray of cups containing oysters and resazurin into incubator in Graham Young’s lab

To speed up sampling and reduce the amount of time cups were out of the incubator, sampling was periodcially divided between two people and two plates (i.e. half the samples collected into plate B, the other half into plate C).

[!NOTE] Based on our experience with resazurin fluorescence stability and the sheer number of samples being handled, resazurin fluorescence was not measured immediately. Plates were measured when time was available, however, no plates sat for longer than 30 minutes at room temperature.

Resazurin temperatures were measured with an infrared thermometer over the course of the experiment:

Timepoint (hrs) Temp_range (C)
0 18
0.5 23-29
1 27-29
1.5 29-31
2 30-32
2.5 31.5-32.5
3 31.5-32.5
3.5 31-32.5
4 31.5-33

Notebook was knitted from 01.00-resazurin-20260713-mgig-36C.Rmd (GitHub).

RESULTS

Family significantly affected size-normalized metabolic rate (AUC ANOVA: F=4.37, p=0.0069), with a significant time × family interaction in the time-series model (LME: p=0.0006). Family 7 had the highest mean metabolic activity and Family 9 the lowest; this contrast was the only significant pairwise difference (Tukey: p=0.0051). The divergence emerged by 1.5 h and strengthened through 4 h (p=0.0002). Families 1 and 5 were intermediate and not significantly different from any other family.

Mean area-under-the-curve (AUC) of blank-corrected metabolism (fold change / mm² shell area) per family across the 4 h heat stress period.
Family n Mean AUC SE
7 20 0.0511 0.0040
1 19 0.0446 0.0049
5 19 0.0373 0.0047
9 20 0.0285 0.0051

Two wells were excluded due to a cup mix-up (samples 70 and 71: one cup contained two oysters, the other none).

AI Use Level 4: Substantial AI contribution

1 Background

To speed up sampling and reduce the amount of time cups were out of the incubator, sampling was periodcially divided between two people and two plates (i.e. half the samples collected into plate B, the other half into plate C).

Based on our experience with resazurin fluorescence stability and the sheer number of samples being handled, resazurin fluorescence was not measured immediately. Plates were measured when time was available, however, no plates sat for longer than 30 minutes at room temperature.

Resazurin temperatures were measured with an infrared thermometer over the course of the experiment:

Timepoint (hrs) Temp_range (C)
0 18
0.5 23-29
1 27-29
1.5 29-31
2 30-32
2.5 31.5-32.5
3 31.5-32.5
3.5 31-32.5
4 31.5-33

See Resazurin/data/20260713-mgig-36C/README.md for full experimental notes.

1.1 Expected inputs

Path Description
Resazurin/data/20260713-mgig-36C/plate-*-T*.txt Plate reader fluorescence exports (one file per plate per timepoint)
Resazurin/data/20260713-mgig-36C/layout.csv Well metadata: plate ID, well ID, blank flag, family groups, sample IDs, area measurements (mm², from ImageJ)

1.2 Expected outputs

All outputs are written to Resazurin/outputs/01.00-resazurin-20260713-mgig-36C/.

File Description
figures/ All plots generated by this script
auc_all_metrics.csv Per-individual AUC values for every active measurement metric
auc_summary.csv Group-level AUC summary statistics (mean, SD, SE, median)
metabolism.csv Full per-well per-timepoint metabolism data frame
pairwise_stats.csv Tukey-adjusted pairwise comparisons from AUC linear models

2 Setup

2.1 Knitr options

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,        # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",         # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Load libraries

library(tidyverse)
library(pracma)       # trapz()
library(lme4)
library(lmerTest)
library(emmeans)
library(multcompView)
library(cowplot)
library(colorspace)   # qualitative_hcl() for large palettes

3 Helper Functions

normalize_well_id <- function(x) {
  x <- toupper(trimws(x))
  valid <- str_detect(x, "^[A-Z]+[0-9]+$")
  out <- rep(NA_character_, length(x))
  if (!any(valid)) return(out)
  m <- str_match(x[valid], "^([A-Z]+)([0-9]+)$")
  out[valid] <- paste0(m[, 2], as.integer(m[, 3]))
  out
}

parse_time_hr <- function(path) {
  hit <- str_match(basename(path),
                   "(?i)-T([0-9]+(?:\\.[0-9]+)?)\\.txt$")
  as.numeric(hit[, 2])
}

parse_plate_id <- function(path) {
  hit <- str_match(basename(path),
    "(?i)^plate-([A-Za-z0-9-]+)-T[0-9]+(?:\\.[0-9]+)?\\.txt$")
  id <- hit[, 2]
  ifelse(is.na(id), "unknown", id)
}

extract_results_block <- function(lines) {
  results_idx <- which(trimws(lines) == "Results")
  if (length(results_idx) == 0) stop("No Results section found")
  idx <- results_idx[1]
  header_tokens <- str_split(lines[idx + 1], "\\t")[[1]] |> trimws()
  col_ids <- header_tokens[
    header_tokens != "" & str_detect(header_tokens, "^[0-9]+$")]
  j <- idx + 2
  data_lines <- character()
  while (j <= length(lines)) {
    line <- lines[j]
    if (trimws(line) == "") break
    if (!str_detect(line, "^[A-Za-z]\\t")) break
    data_lines <- c(data_lines, line)
    j <- j + 1
  }
  list(col_ids = col_ids, data_lines = data_lines)
}

parse_plate_export <- function(path) {
  lines <- readLines(path, warn = FALSE)
  res <- extract_results_block(lines)

  map_dfr(res$data_lines, function(line) {
    tokens <- str_split(line, "\\t")[[1]] |> trimws()
    tokens <- tokens[tokens != ""]
    row_letter <- tokens[1]
    nums <- suppressWarnings(as.numeric(tokens[-1]))
    valid_idx <- which(!is.na(nums))
    if (length(valid_idx) == 0) return(tibble())
    vals <- nums[valid_idx]
    n <- min(length(vals), length(res$col_ids))
    tibble(
      row_id  = toupper(row_letter),
      col_id  = as.integer(res$col_ids[seq_len(n)]),
      well_id = normalize_well_id(
        paste0(toupper(row_letter), res$col_ids[seq_len(n)])),
      value   = vals[seq_len(n)]
    )
  }) %>%
    mutate(
      plate_id = str_to_lower(parse_plate_id(path)),
      time_hr  = parse_time_hr(path)
    )
}

trapezoid_auc <- function(time_hr, value) {
  ok <- is.finite(time_hr) & is.finite(value)
  t <- time_hr[ok]
  v <- value[ok]
  if (length(t) < 2) return(NA_real_)
  ord <- order(t)
  t <- t[ord]; v <- v[ord]
  sum(diff(t) * (head(v, -1) + tail(v, -1)) / 2)
}

# Shared helper: extract display unit string from a measurement column name.
# e.g. "area_mm2_measurement" -> "mm²", "weight_mg_measurement" -> "mg"
parse_meas_unit <- function(col_name) {
  unit_raw <- col_name |>
    str_remove("^metabolism_per_") |>
    str_remove("_measurement$") |>
    str_extract("[^_]+$")
  case_when(
    unit_raw == "mm2" ~ "mm²",
    unit_raw == "cm2" ~ "cm²",
    unit_raw == "mm3" ~ "mm³",
    unit_raw == "cm3" ~ "cm³",
    TRUE              ~ unit_raw
  )
}

# y-axis label for metabolism line plots: "fold change/mm²"
metabolism_y_label <- function(col_name) {
  paste0("Metabolism (fold change/", parse_meas_unit(col_name), ")")
}

# y-axis label for AUC box plots: "Metabolism (AUC; mm²)"
auc_y_label <- function(metric_name) {
  paste0("Metabolism (AUC; ", parse_meas_unit(metric_name), ")")
}

4 Load Data

4.1 Plate export files

proj_root <- rprojroot::find_rstudio_root_file()
data_dir  <- file.path(proj_root, "Resazurin", "data", "20260713-mgig-36C")
out_dir   <- file.path(proj_root, "Resazurin", "outputs",
                        "01.00-resazurin-20260713-mgig-36C")
fig_dir   <- file.path(out_dir, "figures")

dir.create(fig_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)

plate_files <- list.files(
  data_dir,
  pattern = "(?i)^plate-.*-T[0-9]+(?:\\.[0-9]+)?\\.txt$",
  full.names = TRUE
)

plate_raw <- map_dfr(plate_files, function(path) {
  tryCatch(parse_plate_export(path),
           error = function(e) {
             message("Parse error in ", basename(path), ": ", e$message)
             tibble()
           })
})

str(plate_raw)
tibble [1,248 × 6] (S3: tbl_df/tbl/data.frame)
 $ row_id  : chr [1:1248] "A" "A" "A" "A" ...
 $ col_id  : int [1:1248] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id : chr [1:1248] "A1" "A2" "A3" "A4" ...
 $ value   : num [1:1248] 596 597 605 613 601 626 626 626 610 636 ...
 $ plate_id: chr [1:1248] "a" "a" "a" "a" ...
 $ time_hr : num [1:1248] 0 0 0 0 0 0 0 0 0 0 ...

4.2 Plate consistency check

Checks that every plate has the same number of wells at every timepoint. The expected well count is the mode across all plate × timepoint reads. Any plate with at least one deviating read is flagged and dropped entirely before any further analysis — removing only the aberrant timepoint would break the fold-change baseline calculation.

well_counts <- plate_raw %>%
  group_by(plate_id, time_hr) %>%
  summarise(n_wells = n_distinct(well_id), .groups = "drop")

expected_n_wells <- as.integer(
  names(which.max(table(well_counts$n_wells)))
)

inconsistent_reads <- well_counts %>%
  filter(n_wells != expected_n_wells) %>%
  arrange(plate_id, time_hr)

inconsistent_plate_ids <- unique(inconsistent_reads$plate_id)

if (nrow(inconsistent_reads) > 0) {
  cat("**Plate consistency check FAILED.**",
      "Expected", expected_n_wells, "wells per plate-timepoint read.",
      length(inconsistent_plate_ids),
      "plate(s) have at least one deviating read and are excluded",
      "from all analyses:\n\n")
  cat(knitr::kable(
    inconsistent_reads,
    col.names = c("Plate", "Time (h)", "Wells read"),
    caption   = paste("Expected:", expected_n_wells, "wells per read")
  ), sep = "\n")
  cat("\n")
  plate_raw <- plate_raw %>%
    filter(!plate_id %in% inconsistent_plate_ids)
  message(length(inconsistent_plate_ids),
          " plate(s) removed from plate_raw: ",
          paste(inconsistent_plate_ids, collapse = ", "))
} else {
  cat("Plate consistency check passed: all",
      n_distinct(well_counts$plate_id), "plates have",
      expected_n_wells, "wells at every timepoint.\n")
}

Plate consistency check passed: all 3 plates have 96 wells at every timepoint.

4.3 Layout file

layout_path <- file.path(data_dir, "layout.csv")

layout_raw <- read_csv(layout_path,
                       col_types = cols(.default = "c"),
                       show_col_types = FALSE)

# Standardise column names to snake_case
names(layout_raw) <- names(layout_raw) |>
  str_to_lower() |>
  str_replace_all("[^a-z0-9]+", "_") |>
  str_replace_all("_+", "_") |>
  str_replace("_$", "")

# Normalise plate_id to match plate file ids (strip "plate-" prefix)
layout_clean <- layout_raw %>%
  mutate(
    plate_id = str_remove(str_to_lower(plate_id), "^plate-"),
    well_id  = normalize_well_id(plate_well),
    is_blank = if ("is_blank" %in% names(layout_raw))
      toupper(trimws(is_blank)) %in% c("TRUE", "T", "1", "YES", "Y")
    else
      FALSE
  )

found_exclude_col <- intersect(
  c("exclude_from_analysis", "exclude", "omit", "not_analyzed"),
  names(layout_clean)
)[1]
layout_clean <- layout_clean %>%
  mutate(
    exclude_from_analysis = if (!is.na(found_exclude_col))
      toupper(trimws(.data[[found_exclude_col]])) %in%
        c("TRUE", "T", "1", "YES", "Y")
    else
      FALSE
  )

# Identify measurement columns and group columns
measurement_cols <- names(layout_clean)[
  str_detect(names(layout_clean), "_measurement$")]
group_cols <- names(layout_clean)[
  str_detect(names(layout_clean), "_group$")]

# Cast measurement columns to numeric
layout_clean <- layout_clean %>%
  mutate(across(all_of(measurement_cols),
                ~ suppressWarnings(as.numeric(.x))))

# Determine which measurement columns actually contain finite data
active_meas_cols <- measurement_cols[
  sapply(measurement_cols, function(col)
    any(is.finite(layout_clean[[col]]), na.rm = TRUE))]

# Normalise group values to lowercase so they match colour scale definitions
layout_clean <- layout_clean %>%
  mutate(across(all_of(group_cols),
                ~ str_to_lower(trimws(as.character(.x)))))

message("Group columns: ", paste(group_cols, collapse = ", "))
message("Active measurement columns: ",
        paste(active_meas_cols, collapse = ", "))

str(layout_clean)
tibble [235 × 14] (S3: tbl_df/tbl/data.frame)
 $ plate_id             : chr [1:235] "a" "a" "a" "a" ...
 $ plate_well           : chr [1:235] "A01" "A02" "A03" "A04" ...
 $ is_blank             : logi [1:235] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ family_id_group      : chr [1:235] "7" "5" "1" "9" ...
 $ sample_id_group      : chr [1:235] "1" "2" "3" "4" ...
 $ treatment_group      : chr [1:235] NA NA NA NA ...
 $ exclude_from_analysis: logi [1:235] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:235] NA NA NA NA ...
 $ width_mm_measurement : num [1:235] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:235] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:235] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:235] 538 586 713 757 823 ...
 $ imagej_id            : chr [1:235] "2" "4" "1" "1" ...
 $ well_id              : chr [1:235] "A1" "A2" "A3" "A4" ...

5 Merge Plate Data with Layout

dat <- plate_raw %>%
  left_join(
    layout_clean %>%
      select(plate_id, well_id, is_blank, exclude_from_analysis,
             any_of("exclude_reason"),
             all_of(group_cols), all_of(measurement_cols)),
    by = c("plate_id", "well_id")
  ) %>%
  mutate(
    is_blank = replace_na(is_blank, FALSE),
    exclude_from_analysis = replace_na(exclude_from_analysis, FALSE)
  )

str(dat)
tibble [1,248 × 16] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:1248] "A" "A" "A" "A" ...
 $ col_id               : int [1:1248] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id              : chr [1:1248] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:1248] 596 597 605 613 601 626 626 626 610 636 ...
 $ plate_id             : chr [1:1248] "a" "a" "a" "a" ...
 $ time_hr              : num [1:1248] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:1248] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:1248] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:1248] NA NA NA NA ...
 $ family_id_group      : chr [1:1248] "7" "5" "1" "9" ...
 $ sample_id_group      : chr [1:1248] "1" "2" "3" "4" ...
 $ treatment_group      : chr [1:1248] NA NA NA NA ...
 $ width_mm_measurement : num [1:1248] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:1248] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:1248] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:1248] 538 586 713 757 823 ...

6 Raw Fluorescence

6.1 Data frame

# Wells in the plate reader output that have no layout entry get all-NA group
# columns after the join. Keep only wells assigned to at least one group.
active_gc <- intersect(group_cols, names(dat))

raw_df <- dat %>%
  filter(
    !is_blank,
    if (length(active_gc) > 0)
      if_any(all_of(active_gc), ~ !is.na(.))
    else
      TRUE
  ) %>%
  mutate(
    trace_id = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  )

families   <- str_sort(unique(na.omit(raw_df$family_id_group)), numeric = TRUE)
treatments <- sort(unique(na.omit(raw_df$treatment_group)))

n_fam <- length(families)
n_trt <- length(treatments)

# Palette strategy:
#   <= 7 groups : Okabe-Ito (gold standard for colorblind-safe figures).
#   >  7 groups : colorspace::qualitative_hcl("Dynamic") scales to any N
#                 using perceptually uniform HCL space — no colour collisions.
# Black (#000000) is excluded from both and reserved for blank wells.
okabe_ito_7 <- c(
  "#E69F00", "#56B4E9", "#009E73", "#F0E442",
  "#0072B2", "#D55E00", "#CC79A7"
)
make_palette <- function(n) {
  if (n == 0L) return(character(0))
  if (n <= length(okabe_ito_7)) return(okabe_ito_7[seq_len(n)])
  colorspace::qualitative_hcl(n, palette = "Dynamic")
}

all_colours   <- make_palette(n_fam + n_trt)
fam_colours   <- setNames(all_colours[seq_len(n_fam)], families)
trt_colours   <- setNames(all_colours[n_fam + seq_len(n_trt)], treatments)

lty_pool <- c("solid", "dashed", "dotted", "dotdash", "longdash")
trt_linetypes <- setNames(
  lty_pool[(seq_len(n_trt) - 1L) %% length(lty_pool) + 1L],
  treatments
)
plate_well_colours <- c(blank = "black", fam_colours)

has_trt <- n_trt > 0

str(raw_df)
tibble [720 × 17] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:720] "A" "A" "A" "A" ...
 $ col_id               : int [1:720] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id              : chr [1:720] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:720] 596 597 605 613 601 626 626 626 610 636 ...
 $ plate_id             : chr [1:720] "a" "a" "a" "a" ...
 $ time_hr              : num [1:720] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:720] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:720] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:720] NA NA NA NA ...
 $ family_id_group      : chr [1:720] "7" "5" "1" "9" ...
 $ sample_id_group      : chr [1:720] "1" "2" "3" "4" ...
 $ treatment_group      : chr [1:720] NA NA NA NA ...
 $ width_mm_measurement : num [1:720] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:720] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:720] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:720] 538 586 713 757 823 ...
 $ trace_id             : chr [1:720] "1" "2" "3" "4" ...

6.2 Raw fluorescence by plate (including blanks)

p_raw_plates <- dat %>%
  filter(is.finite(time_hr), is.finite(value)) %>%
  mutate(
    colour_group = if_else(is_blank, "blank",
                           coalesce(family_id_group, "sample")),
    trace_id     = paste(plate_id, well_id, sep = "_")
  ) %>%
  ggplot(aes(x = time_hr, y = value,
             group = trace_id, colour = colour_group)) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1, alpha = 0.7) +
  facet_wrap(~ plate_id) +
  scale_colour_manual(
    values   = plate_well_colours,
    name     = "Group",
    breaks   = names(plate_well_colours),
    na.value = "grey80"
  ) +
  labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_raw_plates

ggsave(file.path(fig_dir, "raw_fluor_by_plate.png"),
       p_raw_plates, width = 10, height = 8)

6.3 Mean raw fluorescence by family

raw_family_summary <- raw_df %>%
  filter(!is.na(family_id_group), !exclude_from_analysis) %>%
  group_by(family_id_group, treatment_group, time_hr) %>%
  summarise(
    mean_fluor = mean(value, na.rm = TRUE),
    se_fluor   = sd(value, na.rm = TRUE) /
      sqrt(sum(!is.na(value))),
    n          = sum(!is.na(value)),
    .groups    = "drop"
  ) %>%
  mutate(group_var = if (has_trt)
    paste(family_id_group, treatment_group, sep = ".")
  else
    family_id_group)

p_raw_mean <- ggplot(raw_family_summary,
    aes(x = time_hr, y = mean_fluor,
        colour = family_id_group,
        group = group_var)) +
  geom_ribbon(aes(ymin = mean_fluor - se_fluor,
                  ymax = mean_fluor + se_fluor,
                  fill = family_id_group),
              alpha = 0.15, colour = NA) +
  geom_line(
    mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
    linewidth = 1) +
  geom_point(size = 2) +
  scale_colour_manual(values = fam_colours, name = "Family",
                      breaks = families) +
  scale_fill_manual(values = fam_colours, name = "Family",
                    breaks = families) +
  labs(x = "Time (h)", y = "Mean raw fluorescence (RFU ± SE)") +
  theme_classic(base_size = 13) +
  if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

p_raw_mean

ggsave(file.path(fig_dir, "raw_mean_by_family.png"),
       p_raw_mean, width = 8, height = 5)

6.4 Individual raw fluorescence traces by family

p_raw_by_family <- raw_df %>%
  filter(!is.na(family_id_group)) %>%
  ggplot(aes(x = time_hr, y = value, group = trace_id,
             colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1.2, alpha = 0.7) +
  facet_wrap(~ family_id_group) +
  scale_colour_manual(
    values = if (has_trt) trt_colours else fam_colours,
    name   = if (has_trt) "Treatment" else "Family",
    breaks = if (has_trt) treatments else families) +
  labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_raw_by_family

ggsave(file.path(fig_dir, "raw_individual_by_family.png"),
       p_raw_by_family, width = 10, height = 5)

6.5 Individual raw fluorescence traces by treatment

if (has_trt) {
  p_raw_by_treatment <- raw_df %>%
    ggplot(aes(x = time_hr, y = value,
               group = trace_id, colour = family_id_group)) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ treatment_group) +
    scale_colour_manual(values = fam_colours, name = "Family",
                        breaks = families) +
    labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_raw_by_treatment
  ggsave(file.path(fig_dir, "raw_individual_by_treatment.png"),
         p_raw_by_treatment, width = 10, height = 5)
}

6.6 Excluded samples

Wells flagged exclude_from_analysis = TRUE appear in the raw fluorescence plots above but are omitted from all analyses that follow.

excluded_wells <- dat %>%
  filter(!is_blank, exclude_from_analysis) %>%
  mutate(
    sample = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  ) %>%
  select(plate_id, well_id, sample, family_id_group, treatment_group,
         any_of("exclude_reason")) %>%
  distinct() %>%
  arrange(plate_id, well_id)

if (nrow(excluded_wells) > 0) {
  col_names <- c("Plate", "Well", "Sample", "Family", "Treatment")
  if ("exclude_reason" %in% names(excluded_wells))
    col_names <- c(col_names, "Reason")
  cat(knitr::kable(excluded_wells, col.names = col_names), sep = "\n")
} else {
  cat("No wells are excluded from analysis.\n")
}
Plate Well Sample Family Treatment Reason
a F10 70 5 NA missing oyster - probably second oyster in 71
a F11 71 1 NA two oysters - probably the missing one from 70
b C6 70 5 NA missing oyster - probably second oyster in 71
b C7 71 1 NA two oysters - probably the missing one from 70

7 Blank Correction via Fold-Change Normalization

T0 is the earliest timepoint present in the dataset (not necessarily 0 hr). Sample fold-change is expressed relative to each individual’s T0 reading, resolved by sample_id_group when that column is populated — allowing the same animal to be tracked across plates — or by plate_id + well_id when no sample IDs exist (backward-compatible with single-plate, multi-timepoint designs). Blank fold-change is the per-plate mean blank RFU at each timepoint divided by the pooled mean blank RFU at T0. Subtracting blank fold-change from sample fold-change removes background fluorescence drift; all samples start at exactly 0 at T0 by construction.

7.1 Step 1 – Identify T0 and compute per-sample fold-change

# T0 = earliest timepoint present in the dataset
t0_time <- min(dat$time_hr[is.finite(dat$time_hr)], na.rm = TRUE)
message("T0 timepoint: ", t0_time, " hr")

# T0 reference value per individual.
# Resolved by sample_id_group (cross-plate tracking) when available;
# falls back to plate+well for layouts without explicit sample IDs.
t0_all <- dat %>%
  filter(time_hr == t0_time, !is_blank, is.finite(value)) %>%
  mutate(sample_key = if_else(
    !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
    as.character(sample_id_group),
    paste(plate_id, well_id, sep = "_")
  )) %>%
  group_by(sample_key) %>%
  summarise(value_t0 = mean(value, na.rm = TRUE), .groups = "drop")

dat_fc <- dat %>%
  mutate(sample_key = if_else(
    !is_blank &
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
    as.character(sample_id_group),
    paste(plate_id, well_id, sep = "_")
  )) %>%
  left_join(t0_all, by = "sample_key") %>%
  mutate(fold_change = if_else(
    !is_blank & is.finite(value_t0) & value_t0 > 0,
    value / value_t0,
    NA_real_
  ))

str(dat_fc)
tibble [1,248 × 19] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:1248] "A" "A" "A" "A" ...
 $ col_id               : int [1:1248] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id              : chr [1:1248] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:1248] 596 597 605 613 601 626 626 626 610 636 ...
 $ plate_id             : chr [1:1248] "a" "a" "a" "a" ...
 $ time_hr              : num [1:1248] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:1248] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:1248] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:1248] NA NA NA NA ...
 $ family_id_group      : chr [1:1248] "7" "5" "1" "9" ...
 $ sample_id_group      : chr [1:1248] "1" "2" "3" "4" ...
 $ treatment_group      : chr [1:1248] NA NA NA NA ...
 $ width_mm_measurement : num [1:1248] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:1248] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:1248] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:1248] 538 586 713 757 823 ...
 $ sample_key           : chr [1:1248] "1" "2" "3" "4" ...
 $ value_t0             : num [1:1248] 596 597 605 613 601 626 626 626 610 636 ...
 $ fold_change          : num [1:1248] 1 1 1 1 1 1 1 1 1 1 ...

7.2 Step 2 – Blank fold-change reference per plate per timepoint

# Pooled mean blank RFU at T0 across all T0 plates
mean_blank_t0 <- dat %>%
  filter(is_blank, time_hr == t0_time, is.finite(value)) %>%
  pull(value) %>%
  mean(na.rm = TRUE)

if (!is.finite(mean_blank_t0))
  message("No blank readings found at T0 (", t0_time,
          " hr); blank correction will produce NA.")

# Per-plate per-timepoint mean blank expressed as fold-change relative to T0
blank_fc_ref <- dat %>%
  filter(is_blank, is.finite(value)) %>%
  group_by(plate_id, time_hr) %>%
  summarise(mean_blank_rfu = mean(value, na.rm = TRUE), .groups = "drop") %>%
  mutate(mean_blank_fc = mean_blank_rfu / mean_blank_t0)

str(blank_fc_ref)
tibble [13 × 4] (S3: tbl_df/tbl/data.frame)
 $ plate_id      : chr [1:13] "a" "a" "a" "a" ...
 $ time_hr       : num [1:13] 0 2.5 3 3.5 4 0.5 1 1.5 2 0.5 ...
 $ mean_blank_rfu: num [1:13] 663 731 711 704 722 ...
 $ mean_blank_fc : num [1:13] 1 1.1 1.07 1.06 1.09 ...

7.3 Step 3 – Subtract blank fold-change from sample fold-change

samples <- dat_fc %>%
  filter(!is_blank, !exclude_from_analysis) %>%
  mutate(
    trace_id = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  ) %>%
  left_join(blank_fc_ref, by = c("plate_id", "time_hr")) %>%
  mutate(corrected_fc = fold_change - mean_blank_fc)

str(samples)
tibble [1,191 × 23] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:1191] "A" "A" "A" "A" ...
 $ col_id               : int [1:1191] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id              : chr [1:1191] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:1191] 596 597 605 613 601 626 626 626 610 636 ...
 $ plate_id             : chr [1:1191] "a" "a" "a" "a" ...
 $ time_hr              : num [1:1191] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:1191] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:1191] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:1191] NA NA NA NA ...
 $ family_id_group      : chr [1:1191] "7" "5" "1" "9" ...
 $ sample_id_group      : chr [1:1191] "1" "2" "3" "4" ...
 $ treatment_group      : chr [1:1191] NA NA NA NA ...
 $ width_mm_measurement : num [1:1191] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:1191] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement: num [1:1191] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:1191] 538 586 713 757 823 ...
 $ sample_key           : chr [1:1191] "1" "2" "3" "4" ...
 $ value_t0             : num [1:1191] 596 597 605 613 601 626 626 626 610 636 ...
 $ fold_change          : num [1:1191] 1 1 1 1 1 1 1 1 1 1 ...
 $ trace_id             : chr [1:1191] "1" "2" "3" "4" ...
 $ mean_blank_rfu       : num [1:1191] 663 663 663 663 663 ...
 $ mean_blank_fc        : num [1:1191] 1 1 1 1 1 1 1 1 1 1 ...
 $ corrected_fc         : num [1:1191] 0 0 0 0 0 0 0 0 0 0 ...

8 Blank-Corrected Fold-Change

8.1 Mean by family

bc_fc_summary <- samples %>%
  filter(!is.na(family_id_group), !exclude_from_analysis) %>%
  group_by(family_id_group, treatment_group, time_hr) %>%
  summarise(
    mean_val = mean(corrected_fc, na.rm = TRUE),
    se_val   = sd(corrected_fc, na.rm = TRUE) /
      sqrt(sum(!is.na(corrected_fc))),
    n        = sum(!is.na(corrected_fc)),
    .groups  = "drop"
  ) %>%
  mutate(group_var = if (has_trt)
    paste(family_id_group, treatment_group, sep = ".")
  else
    family_id_group)

p_bc_fc_mean <- ggplot(bc_fc_summary,
    aes(x = time_hr, y = mean_val,
        colour = family_id_group,
        group = group_var)) +
  geom_ribbon(aes(ymin = mean_val - se_val,
                  ymax = mean_val + se_val,
                  fill = family_id_group),
              alpha = 0.15, colour = NA) +
  geom_line(
    mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
    linewidth = 1) +
  geom_point(size = 2) +
  scale_colour_manual(values = fam_colours, name = "Family",
                      breaks = families) +
  scale_fill_manual(values = fam_colours, name = "Family",
                    breaks = families) +
  labs(x = "Time (h)",
       y = "Mean blank-corrected fold-change (± SE)") +
  theme_classic(base_size = 13) +
  if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

p_bc_fc_mean

ggsave(file.path(fig_dir, "blank_corrected_fc_mean_by_family.png"),
       p_bc_fc_mean, width = 8, height = 5)

8.2 Individual traces by family

p_bc_fc_by_family <- samples %>%
  filter(!is.na(family_id_group)) %>%
  ggplot(aes(x = time_hr, y = corrected_fc, group = trace_id,
             colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1.2, alpha = 0.7) +
  facet_wrap(~ family_id_group) +
  scale_colour_manual(
    values = if (has_trt) trt_colours else fam_colours,
    name   = if (has_trt) "Treatment" else "Family",
    breaks = if (has_trt) treatments else families) +
  labs(x = "Time (h)", y = "Blank-corrected fold-change") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_bc_fc_by_family

ggsave(file.path(fig_dir, "blank_corrected_fc_by_family.png"),
       p_bc_fc_by_family, width = 10, height = 5)

8.3 Individual blank-corrected fold-change traces by treatment

if (has_trt) {
  p_bc_fc_by_treatment <- samples %>%
    ggplot(aes(x = time_hr, y = corrected_fc,
               group = trace_id, colour = family_id_group)) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ treatment_group) +
    scale_colour_manual(values = fam_colours, name = "Family",
                        breaks = families) +
    labs(x = "Time (h)", y = "Blank-corrected fold-change") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_bc_fc_by_treatment
  ggsave(file.path(fig_dir, "blank_corrected_fc_by_treatment.png"),
         p_bc_fc_by_treatment, width = 10, height = 5)
}

9 Metabolism (Size-Normalised Fold-Change)

Blank-corrected fold-change divided by each active measurement column. This is “metabolism” as defined in Huffmyer et al.

if (length(active_meas_cols) == 0) {
  message("No active measurement columns: skipping metabolism calculation.")
  metabolism_df <- tibble()
} else {
  metabolism_df <- samples
  for (mc in active_meas_cols) {
    out_col <- paste0("metabolism_per_", mc)
    metabolism_df <- metabolism_df %>%
      mutate(!!out_col := if_else(
        is.finite(.data[[mc]]) & .data[[mc]] > 0 &
          is.finite(corrected_fc),
        corrected_fc / .data[[mc]],
        NA_real_
      ))
  }
}

str(metabolism_df)
tibble [1,191 × 24] (S3: tbl_df/tbl/data.frame)
 $ row_id                             : chr [1:1191] "A" "A" "A" "A" ...
 $ col_id                             : int [1:1191] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id                            : chr [1:1191] "A1" "A2" "A3" "A4" ...
 $ value                              : num [1:1191] 596 597 605 613 601 626 626 626 610 636 ...
 $ plate_id                           : chr [1:1191] "a" "a" "a" "a" ...
 $ time_hr                            : num [1:1191] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank                           : logi [1:1191] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis              : logi [1:1191] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason                     : chr [1:1191] NA NA NA NA ...
 $ family_id_group                    : chr [1:1191] "7" "5" "1" "9" ...
 $ sample_id_group                    : chr [1:1191] "1" "2" "3" "4" ...
 $ treatment_group                    : chr [1:1191] NA NA NA NA ...
 $ width_mm_measurement               : num [1:1191] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement              : num [1:1191] NA NA NA NA NA NA NA NA NA NA ...
 $ weight_mg_measurement              : num [1:1191] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement               : num [1:1191] 538 586 713 757 823 ...
 $ sample_key                         : chr [1:1191] "1" "2" "3" "4" ...
 $ value_t0                           : num [1:1191] 596 597 605 613 601 626 626 626 610 636 ...
 $ fold_change                        : num [1:1191] 1 1 1 1 1 1 1 1 1 1 ...
 $ trace_id                           : chr [1:1191] "1" "2" "3" "4" ...
 $ mean_blank_rfu                     : num [1:1191] 663 663 663 663 663 ...
 $ mean_blank_fc                      : num [1:1191] 1 1 1 1 1 1 1 1 1 1 ...
 $ corrected_fc                       : num [1:1191] 0 0 0 0 0 0 0 0 0 0 ...
 $ metabolism_per_area_mm2_measurement: num [1:1191] 0 0 0 0 0 0 0 0 0 0 ...

9.1 Mean metabolism by family

if (nrow(metabolism_df) > 0) {

  metab_cols <- paste0("metabolism_per_", active_meas_cols)

  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    metab_summary <- metabolism_df %>%
      filter(!is.na(family_id_group), !exclude_from_analysis) %>%
      group_by(family_id_group, treatment_group, time_hr) %>%
      summarise(
        mean_val = mean(.data[[col]], na.rm = TRUE),
        se_val   = sd(.data[[col]], na.rm = TRUE) /
          sqrt(sum(!is.na(.data[[col]]))),
        .groups  = "drop"
      ) %>%
      mutate(group_var = if (has_trt)
        paste(family_id_group, treatment_group, sep = ".")
      else
        family_id_group)

    p_metab_mean <- ggplot(metab_summary,
        aes(x = time_hr, y = mean_val,
            colour = family_id_group,
            group = group_var)) +
      geom_ribbon(aes(ymin = mean_val - se_val,
                      ymax = mean_val + se_val,
                      fill = family_id_group),
                  alpha = 0.15, colour = NA) +
      geom_line(
        mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
        linewidth = 1) +
      geom_point(size = 2) +
      scale_colour_manual(values = fam_colours, name = "Family",
                          breaks = families) +
      scale_fill_manual(values = fam_colours, name = "Family",
                        breaks = families) +
      labs(x = "Time (h)",
           y = paste0(metabolism_y_label(col), " (± SE)")) +
      theme_classic(base_size = 13) +
      if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

    print(p_metab_mean)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_mean_", mc_label, ".png")),
      p_metab_mean, width = 8, height = 5)
  }
}

9.2 Individual metabolism traces by family

if (nrow(metabolism_df) > 0) {

  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    p_metab_by_family <- metabolism_df %>%
      filter(!is.na(family_id_group)) %>%
      ggplot(aes(x = time_hr, y = .data[[col]], group = trace_id,
                 colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
      geom_line(alpha = 0.6) +
      geom_point(size = 1.2, alpha = 0.7) +
      facet_wrap(~ family_id_group) +
      scale_colour_manual(
        values = if (has_trt) trt_colours else fam_colours,
        name   = if (has_trt) "Treatment" else "Family",
        breaks = if (has_trt) treatments else families) +
      labs(x = "Time (h)", y = metabolism_y_label(col)) +
      theme_classic(base_size = 12) +
      theme(strip.background = element_blank(),
            strip.text       = element_text(face = "bold"))

    print(p_metab_by_family)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_individual_", mc_label, "_by_family.png")),
      p_metab_by_family, width = 10, height = 5)

    if (has_trt) {
      p_metab_by_treatment <- ggplot(metabolism_df,
          aes(x = time_hr, y = .data[[col]],
              group = trace_id, colour = family_id_group)) +
        geom_line(alpha = 0.6) +
        geom_point(size = 1.2, alpha = 0.7) +
        facet_wrap(~ treatment_group) +
        scale_colour_manual(values = fam_colours, name = "Family",
                            breaks = families) +
        labs(x = "Time (h)", y = metabolism_y_label(col)) +
        theme_classic(base_size = 12) +
        theme(strip.background = element_blank(),
              strip.text       = element_text(face = "bold"))

      print(p_metab_by_treatment)
      ggsave(
        file.path(fig_dir,
                  paste0("metabolism_individual_", mc_label, "_by_treatment.png")),
        p_metab_by_treatment, width = 10, height = 5)
    }
  }
}

10 Time-Series Statistical Analysis

Linear mixed effects models test the effect of experimental variables on metabolism over time. Individual (sample_id_group) is included as a random intercept to account for repeated measures across timepoints. Type III ANOVA with Satterthwaite’s approximation (lmerTest) assesses significance; post-hoc pairwise comparisons use estimated marginal means (emmeans, Tukey adjustment).

run_ts_stats <- function(df, value_col) {
  has_family    <- "family_id_group" %in% names(df) &&
    length(unique(na.omit(df$family_id_group))) > 1
  has_treatment <- "treatment_group" %in% names(df) &&
    length(unique(na.omit(df$treatment_group))) > 1

  if (!has_family && !has_treatment) return(NULL)

  df <- df %>%
    filter(is.finite(.data[[value_col]]), is.finite(time_hr)) %>%
    mutate(
      time_f     = factor(time_hr),
      individual = factor(trace_id)
    )

  if (nrow(df) == 0) return(NULL)

  if (has_family)    df <- df %>% mutate(family    = factor(family_id_group))
  if (has_treatment) df <- df %>% mutate(treatment = factor(treatment_group))

  if (has_family    && length(unique(na.omit(df$family)))    < 2) return(NULL)
  if (has_treatment && length(unique(na.omit(df$treatment))) < 2) return(NULL)

  fixed <- if (has_family && has_treatment) {
    paste0(value_col, " ~ time_f * family * treatment")
  } else if (has_family) {
    paste0(value_col, " ~ time_f * family")
  } else {
    paste0(value_col, " ~ time_f * treatment")
  }

  model <- lmer(
    as.formula(paste0(fixed, " + (1 | individual)")),
    data = df
  )

  anova_res <- anova(model, type = 3, ddf = "Satterthwaite")

  # Pairwise comparisons of group combinations at each timepoint
  emm_spec <- if (has_family && has_treatment) {
    ~ family * treatment | time_f
  } else if (has_family) {
    ~ family | time_f
  } else {
    ~ treatment | time_f
  }

  emm       <- emmeans(model, emm_spec)
  pairs_res <- as.data.frame(pairs(emm, adjust = "tukey"))

  # Main-effect marginal means (collapsed across time)
  emm_main <- if (has_family && has_treatment) {
    emmeans(model, ~ family * treatment)
  } else if (has_family) {
    emmeans(model, ~ family)
  } else {
    emmeans(model, ~ treatment)
  }

  pairs_main <- as.data.frame(pairs(emm_main, adjust = "tukey"))

  list(
    model         = model,
    anova         = anova_res,
    pairs_by_time = pairs_res,
    pairs_main    = pairs_main,
    has_family    = has_family,
    has_treatment = has_treatment
  )
}

ts_stats <- list()
if (nrow(metabolism_df) > 0) {
  for (mc in active_meas_cols) {
    col <- paste0("metabolism_per_", mc)
    if (col %in% names(metabolism_df))
      ts_stats[[col]] <- run_ts_stats(metabolism_df, col)
  }
}

10.1 Results

for (col in names(ts_stats)) {
  res <- ts_stats[[col]]
  if (is.null(res)) next

  cat("\n\n### Metric:", col, "\n\n")

  cat("**Type III ANOVA (Satterthwaite approximation):**\n\n")
  cat(knitr::kable(as.data.frame(res$anova), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")

  cat("**Marginal means – main effects (collapsed across time):**\n\n")
  cat(knitr::kable(as.data.frame(res$pairs_main), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")

  cat("**Pairwise comparisons by timepoint (Tukey):**\n\n")
  cat(knitr::kable(as.data.frame(res$pairs_by_time), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
}

10.1.1 Metric: metabolism_per_area_mm2_measurement

Type III ANOVA (Satterthwaite approximation):

Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
time_f 0.0269 0.0034 8 592 201.3951 0.0000
family 0.0002 0.0001 3 74 4.1895 0.0085
time_f:family 0.0009 0.0000 24 592 2.2689 0.0006

Marginal means – main effects (collapsed across time):

contrast estimate SE df t.ratio p.value
family1 - family5 0.0018 0.0017 74 1.0878 0.6980
family1 - family7 -0.0015 0.0017 74 -0.9143 0.7973
family1 - family9 0.0040 0.0017 74 2.4052 0.0849
family5 - family7 -0.0033 0.0017 74 -2.0160 0.1914
family5 - family9 0.0022 0.0017 74 1.3035 0.5636
family7 - family9 0.0055 0.0016 74 3.3629 0.0066

Pairwise comparisons by timepoint (Tukey):

contrast time_f estimate SE df t.ratio p.value
family1 - family5 0 0.0000 0.0021 172.1828 0.0000 1.0000
family1 - family7 0 0.0000 0.0021 172.1828 0.0000 1.0000
family1 - family9 0 0.0000 0.0021 172.1828 0.0000 1.0000
family5 - family7 0 0.0000 0.0021 172.1828 0.0000 1.0000
family5 - family9 0 0.0000 0.0021 172.1828 0.0000 1.0000
family7 - family9 0 0.0000 0.0020 172.1828 0.0000 1.0000
family1 - family5 0.5 -0.0004 0.0021 172.1828 -0.1799 0.9979
family1 - family7 0.5 -0.0015 0.0021 172.1828 -0.7412 0.8803
family1 - family9 0.5 0.0007 0.0021 172.1828 0.3484 0.9854
family5 - family7 0.5 -0.0012 0.0021 172.1828 -0.5591 0.9440
family5 - family9 0.5 0.0011 0.0021 172.1828 0.5306 0.9515
family7 - family9 0.5 0.0023 0.0020 172.1828 1.1039 0.6876
family1 - family5 1 0.0001 0.0021 172.1828 0.0474 1.0000
family1 - family7 1 -0.0029 0.0021 172.1828 -1.4243 0.4860
family1 - family9 1 0.0019 0.0021 172.1828 0.9313 0.7881
family5 - family7 1 -0.0030 0.0021 172.1828 -1.4723 0.4565
family5 - family9 1 0.0018 0.0021 172.1828 0.8833 0.8135
family7 - family9 1 0.0049 0.0020 172.1828 2.3864 0.0835
family1 - family5 1.5 0.0014 0.0021 172.1828 0.6526 0.9145
family1 - family7 1.5 -0.0023 0.0021 172.1828 -1.1083 0.6849
family1 - family9 1.5 0.0030 0.0021 172.1828 1.4675 0.4594
family5 - family7 1.5 -0.0037 0.0021 172.1828 -1.7692 0.2917
family5 - family9 1.5 0.0017 0.0021 172.1828 0.8066 0.8512
family7 - family9 1.5 0.0053 0.0020 172.1828 2.6095 0.0481
family1 - family5 2 0.0023 0.0021 172.1828 1.0860 0.6985
family1 - family7 2 -0.0018 0.0021 172.1828 -0.8513 0.8297
family1 - family9 2 0.0041 0.0021 172.1828 1.9992 0.1923
family5 - family7 2 -0.0040 0.0021 172.1828 -1.9512 0.2108
family5 - family9 2 0.0019 0.0021 172.1828 0.8994 0.8052
family7 - family9 2 0.0059 0.0020 172.1828 2.8878 0.0225
family1 - family5 2.5 0.0032 0.0021 172.1828 1.5385 0.4168
family1 - family7 2.5 -0.0013 0.0021 172.1828 -0.6306 0.9221
family1 - family9 2.5 0.0056 0.0021 172.1828 2.7107 0.0368
family5 - family7 2.5 -0.0045 0.0021 172.1828 -2.1887 0.1305
family5 - family9 2.5 0.0024 0.0021 172.1828 1.1526 0.6575
family7 - family9 2.5 0.0069 0.0020 172.1828 3.3850 0.0048
family1 - family5 3 0.0031 0.0021 172.1828 1.4628 0.4623
family1 - family7 3 -0.0011 0.0021 172.1828 -0.5329 0.9510
family1 - family9 3 0.0062 0.0021 172.1828 3.0222 0.0152
family5 - family7 3 -0.0042 0.0021 172.1828 -2.0143 0.1867
family5 - family9 3 0.0032 0.0021 172.1828 1.5408 0.4155
family7 - family9 3 0.0073 0.0020 172.1828 3.6016 0.0023
family1 - family5 3.5 0.0031 0.0021 172.1828 1.4823 0.4504
family1 - family7 3.5 -0.0015 0.0021 172.1828 -0.7074 0.8940
family1 - family9 3.5 0.0067 0.0021 172.1828 3.2489 0.0075
family5 - family7 3.5 -0.0046 0.0021 172.1828 -2.2086 0.1250
family5 - family9 3.5 0.0036 0.0021 172.1828 1.7477 0.3024
family7 - family9 3.5 0.0082 0.0020 172.1828 4.0080 0.0005
family1 - family5 4 0.0037 0.0021 172.1828 1.7631 0.2948
family1 - family7 4 -0.0012 0.0021 172.1828 -0.6040 0.9307
family1 - family9 4 0.0075 0.0021 172.1828 3.6345 0.0021
family5 - family7 4 -0.0049 0.0021 172.1828 -2.3895 0.0829
family5 - family9 4 0.0038 0.0021 172.1828 1.8489 0.2542
family7 - family9 4 0.0088 0.0020 172.1828 4.2938 0.0002

11 Area Under the Curve (AUC)

AUC computed per individual via the trapezoid rule across all timepoints. metabolism_per_* is the primary metric matching the paper; corrected_fc and raw_fluorescence are retained for reference.

compute_auc <- function(df, value_col, group_vars) {
  df %>%
    filter(is.finite(time_hr), is.finite(.data[[value_col]])) %>%
    group_by(across(all_of(group_vars))) %>%
    summarise(
      AUC          = trapezoid_auc(time_hr, .data[[value_col]]),
      n_timepoints = n(),
      .groups      = "drop"
    ) %>%
    filter(is.finite(AUC))
}

# Only include grouping columns that are actually present in the data
individual_vars <- intersect(
  c("trace_id", "family_id_group", "treatment_group"),
  names(metabolism_df)
)

auc_metab_list <- list()
if (nrow(metabolism_df) > 0) {
  for (mc in active_meas_cols) {
    col <- paste0("metabolism_per_", mc)
    if (col %in% names(metabolism_df)) {
      auc_metab_list[[col]] <-
        compute_auc(metabolism_df, col, individual_vars) %>%
        mutate(metric = col)
    }
  }
}

auc_all <- bind_rows(auc_metab_list)

str(auc_all)
tibble [78 × 6] (S3: tbl_df/tbl/data.frame)
 $ trace_id       : chr [1:78] "1" "10" "11" "12" ...
 $ family_id_group: chr [1:78] "7" "5" "1" "9" ...
 $ treatment_group: chr [1:78] NA NA NA NA ...
 $ AUC            : num [1:78] 0.0359 0.0813 0.0489 0.0122 0.0505 ...
 $ n_timepoints   : int [1:78] 9 9 9 9 9 9 9 9 9 9 ...
 $ metric         : chr [1:78] "metabolism_per_area_mm2_measurement" "metabolism_per_area_mm2_measurement" "metabolism_per_area_mm2_measurement" "metabolism_per_area_mm2_measurement" ...

11.1 AUC summary tables

sum_vars <- intersect(
  c("metric", "family_id_group", "treatment_group"),
  names(auc_all)
)
auc_summary <- auc_all %>%
  group_by(across(all_of(sum_vars))) %>%
  summarise(
    n      = n(),
    mean   = mean(AUC, na.rm = TRUE),
    sd     = sd(AUC, na.rm = TRUE),
    se     = sd / sqrt(n),
    median = median(AUC, na.rm = TRUE),
    .groups = "drop"
  )

print(auc_summary)
# A tibble: 4 × 8
  metric      family_id_group treatment_group     n   mean     sd      se median
  <chr>       <chr>           <chr>           <int>  <dbl>  <dbl>   <dbl>  <dbl>
1 metabolism… 1               <NA>               19 0.0446 0.0215 0.00493 0.0441
2 metabolism… 5               <NA>               19 0.0373 0.0204 0.00467 0.0366
3 metabolism… 7               <NA>               20 0.0511 0.0179 0.00400 0.0513
4 metabolism… 9               <NA>               20 0.0285 0.0228 0.00510 0.0124

12 Statistical Analysis

Each individual oyster (sample_id_group) is the observational unit. The model is built from whichever grouping factors are present: both family and treatment (with interaction) when both exist, or a one-way model when only one factor is available. Each plate maps to a unique family × treatment combination, so plate-level and group-level variance are confounded; interpret accordingly.

run_auc_stats <- function(auc_df) {
  empty <- tibble()

  has_family    <- "family_id_group" %in% names(auc_df) &&
    length(unique(na.omit(auc_df$family_id_group))) > 1
  has_treatment <- "treatment_group" %in% names(auc_df) &&
    length(unique(na.omit(auc_df$treatment_group))) > 1

  if (!has_family && !has_treatment) {
    return(list(model = NULL, anova = NULL,
                pairs_full = empty, pairs_family = empty,
                pairs_trt = empty,
                has_family = FALSE, has_treatment = FALSE))
  }

  if (has_family)    auc_df <- auc_df %>% mutate(family    = factor(family_id_group))
  if (has_treatment) auc_df <- auc_df %>% mutate(treatment = factor(treatment_group))

  formula_str <- if (has_family && has_treatment) {
    "AUC ~ family * treatment"
  } else if (has_family) {
    "AUC ~ family"
  } else {
    "AUC ~ treatment"
  }
  model     <- lm(as.formula(formula_str), data = auc_df)
  anova_res <- anova(model)

  if (has_family && has_treatment) {
    pairs_full   <- as.data.frame(pairs(emmeans(model, ~ family * treatment),
                                        adjust = "tukey"))
    pairs_family <- as.data.frame(pairs(emmeans(model, ~ family),
                                        adjust = "tukey"))
    pairs_trt    <- as.data.frame(pairs(emmeans(model, ~ treatment),
                                        adjust = "tukey"))
  } else if (has_family) {
    pairs_family <- as.data.frame(pairs(emmeans(model, ~ family),
                                        adjust = "tukey"))
    pairs_full   <- pairs_family
    pairs_trt    <- empty
  } else {
    pairs_trt    <- as.data.frame(pairs(emmeans(model, ~ treatment),
                                        adjust = "tukey"))
    pairs_full   <- pairs_trt
    pairs_family <- empty
  }

  list(
    model         = model,
    anova         = anova_res,
    pairs_full    = pairs_full,
    pairs_family  = pairs_family,
    pairs_trt     = pairs_trt,
    has_family    = has_family,
    has_treatment = has_treatment
  )
}

metrics_to_test <- unique(auc_all$metric)
stats_results   <- map(
  set_names(metrics_to_test),
  ~ run_auc_stats(auc_all %>% filter(metric == .x))
)

12.1 Results by metric

for (met in metrics_to_test) {
  stats <- stats_results[[met]]
  cat("\n\n### Metric:", met, "\n\n")
  cat("**ANOVA:**\n\n")
  cat(knitr::kable(as.data.frame(stats$anova), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  if (stats$has_family && stats$has_treatment) {
    cat("**Pairwise: family × treatment (Tukey):**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_full), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
    cat("**Pairwise: family main effect:**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_family), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
    cat("**Pairwise: treatment main effect:**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_trt), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  } else if (stats$has_family) {
    cat("**Pairwise: family (Tukey):**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_family), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  } else if (stats$has_treatment) {
    cat("**Pairwise: treatment (Tukey):**\n\n")
    cat(knitr::kable(as.data.frame(stats$pairs_trt), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
  }
}

12.1.1 Metric: metabolism_per_area_mm2_measurement

ANOVA:

Df Sum Sq Mean Sq F value Pr(>F)
family 3 0.0056 0.0019 4.3721 0.0069
Residuals 74 0.0317 0.0004 NA NA

Pairwise: family (Tukey):

contrast estimate SE df t.ratio p.value
family1 - family5 0.0073 0.0067 74 1.0858 0.6993
family1 - family7 -0.0065 0.0066 74 -0.9809 0.7607
family1 - family9 0.0161 0.0066 74 2.4212 0.0819
family5 - family7 -0.0138 0.0066 74 -2.0806 0.1690
family5 - family9 0.0088 0.0066 74 1.3216 0.5523
family7 - family9 0.0226 0.0065 74 3.4467 0.0051

13 AUC Box Plots with Statistical Annotations

Significance labels: *** p < 0.001, ** p < 0.01, * p < 0.05. Brackets are drawn only for significant pairs (p < 0.05). Plots are generated for whichever grouping factors are present: treatment-only, family-only, all-groups, within-family, and within-treatment.

sig_label <- function(p) {
  case_when(p < 0.001 ~ "***", p < 0.01 ~ "**", p < 0.05 ~ "*",
            TRUE ~ "ns")
}

# Add significance brackets to an existing ggplot.
# pairs_df   : data frame with $contrast and $p.value columns
# group_levels: ordered character vector matching x-axis factor levels
# y_vals     : numeric vector of AUC values used to set bracket heights
add_sig_brackets <- function(p, pairs_df, group_levels, y_vals) {
  sig_pairs <- pairs_df %>%
    mutate(label = sig_label(p.value)) %>%
    filter(label != "ns")
  if (nrow(sig_pairs) == 0) return(p)

  y_max   <- max(y_vals, na.rm = TRUE)
  y_range <- diff(range(y_vals, na.rm = TRUE))
  step    <- y_range * 0.12

  # emmeans prefixes factor levels with the variable name (e.g. "family7");
  # resolve back to the bare level used on the x-axis.
  resolve_level <- function(x, lvls) {
    x <- trimws(x)
    if (x %in% lvls) return(x)
    lvls_sorted <- lvls[order(nchar(as.character(lvls)), decreasing = TRUE)]
    for (lvl in lvls_sorted) {
      lvl_str <- as.character(lvl)
      if (endsWith(x, lvl_str) && nchar(x) > nchar(lvl_str)) return(lvl_str)
    }
    NA_character_
  }

  for (i in seq_len(nrow(sig_pairs))) {
    parts <- str_split(as.character(sig_pairs$contrast[i]), " - ", 2)[[1]]
    g1 <- resolve_level(parts[1], group_levels)
    g2 <- resolve_level(parts[2], group_levels)
    x1 <- match(g1, group_levels)
    x2 <- match(g2, group_levels)
    if (is.na(x1) || is.na(x2)) next
    bar_y <- y_max + i * step
    p <- p +
      annotate("segment", x = x1, xend = x2,
               y = bar_y, yend = bar_y,
               colour = "black", linewidth = 0.6) +
      annotate("segment", x = x1, xend = x1,
               y = bar_y, yend = bar_y - step * 0.3,
               colour = "black", linewidth = 0.6) +
      annotate("segment", x = x2, xend = x2,
               y = bar_y, yend = bar_y - step * 0.3,
               colour = "black", linewidth = 0.6) +
      annotate("text", x = (x1 + x2) / 2,
               y = bar_y + step * 0.15,
               label = sig_pairs$label[i], size = 4.5)
  }
  p
}
for (met in metrics_to_test) {
  df      <- auc_all %>% filter(metric == met)
  stats   <- stats_results[[met]]
  y_lab   <- auc_y_label(met)
  has_fam <- stats$has_family
  has_trt <- stats$has_treatment

  # ── Treatment main effect (x = treatment, tick = treatment name) ───────
  if (has_trt) {
    df_p <- df %>%
      mutate(x = factor(treatment_group, levels = sort(unique(treatment_group))))
    grps <- levels(df_p$x)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = trt_colours[grps], guide = "none") +
      labs(x = "Treatment", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, stats$pairs_trt, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_treatment_", met, ".png")),
           p, width = 5, height = 5)
  }

  # ── Family main effect (x = family, tick = family name) ───────────────
  if (has_fam) {
    df_p <- df %>%
      mutate(x = factor(family_id_group,
                        levels = str_sort(unique(family_id_group), numeric = TRUE)))
    grps <- levels(df_p$x)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = fam_colours[grps], guide = "none") +
      labs(x = "Family", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, stats$pairs_family, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_family_", met, ".png")),
           p, width = 5, height = 5)
  }

  # Remaining plots require both factors
  if (!has_fam || !has_trt) next

  # ── All family:treatment groups (x = family:treatment) ─────────────────
  # emmeans contrasts use spaces; convert to colon to match tick labels
  pairs_fc <- stats$pairs_full %>%
    mutate(contrast = str_replace_all(
      contrast,
      "([a-z]+) ([a-z]+)( - )([a-z]+) ([a-z]+)",
      "\\1:\\2\\3\\4:\\5"
    ))
  df_p <- df %>%
    mutate(x = factor(
      paste(family_id_group, treatment_group, sep = ":"),
      levels = str_sort(unique(paste(family_id_group, treatment_group, sep = ":")),
                        numeric = TRUE)
    ))
  grps     <- levels(df_p$x)
  fill_map <- setNames(make_palette(length(grps)), grps)
  p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
    geom_boxplot(alpha = 0.6, outlier.shape = NA) +
    geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
    scale_fill_manual(values = fill_map, guide = "none") +
    labs(x = "Family : Treatment", y = y_lab) +
    theme_classic(base_size = 13) +
    theme(axis.text.x = element_text(angle = 20, hjust = 1))
  p <- add_sig_brackets(p, pairs_fc, grps, df_p$AUC)
  print(p)
  ggsave(file.path(fig_dir, paste0("auc_all_groups_", met, ".png")),
         p, width = 6, height = 5)

  # ── Within each family: treatment comparison (x = family:treatment) ────
  # Tick labels are family:treatment so these plots are visually distinct
  # from the treatment main-effect plot above.
  for (fam in str_sort(unique(df$family_id_group), numeric = TRUE)) {
    df_p <- df %>%
      filter(family_id_group == fam) %>%
      mutate(x = factor(
        paste(family_id_group, treatment_group, sep = ":"),
        levels = str_sort(unique(paste(family_id_group, treatment_group, sep = ":")),
                          numeric = TRUE)
      ))
    grps     <- levels(df_p$x)
    pairs_sub <- pairs_fc %>%
      filter(str_count(contrast, paste0(fam, ":")) == 2)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = fill_map[grps], guide = "none") +
      labs(x = "Family : Treatment", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, pairs_sub, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_", fam, "_trt_", met, ".png")),
           p, width = 5, height = 5)
  }

  # ── Within each treatment: family comparison (x = family:treatment) ────
  # Tick labels are family:treatment so these plots are visually distinct
  # from the family main-effect plot above.
  for (trt in sort(unique(df$treatment_group))) {
    df_p <- df %>%
      filter(treatment_group == trt) %>%
      mutate(x = factor(
        paste(family_id_group, treatment_group, sep = ":"),
        levels = str_sort(unique(paste(family_id_group, treatment_group, sep = ":")),
                          numeric = TRUE)
      ))
    grps      <- levels(df_p$x)
    pairs_sub <- pairs_fc %>%
      filter(str_count(contrast, paste0(":", trt)) == 2)
    p <- ggplot(df_p, aes(x = x, y = AUC, fill = x)) +
      geom_boxplot(alpha = 0.6, outlier.shape = NA) +
      geom_jitter(width = 0.15, alpha = 0.4, size = 1.5) +
      scale_fill_manual(values = fill_map[grps], guide = "none") +
      labs(x = "Family : Treatment", y = y_lab) +
      theme_classic(base_size = 13)
    p <- add_sig_brackets(p, pairs_sub, grps, df_p$AUC)
    print(p)
    ggsave(file.path(fig_dir, paste0("auc_", trt, "_fam_", met, ".png")),
           p, width = 5, height = 5)
  }
}

14 Save Output Data

write_csv(auc_all,      file.path(out_dir, "auc_all_metrics.csv"))
write_csv(auc_summary,  file.path(out_dir, "auc_summary.csv"))

if (nrow(metabolism_df) > 0)
  write_csv(metabolism_df,
            file.path(out_dir, "metabolism.csv"))

stats_compiled <- map_dfr(metrics_to_test, function(met) {
  bind_rows(
    stats_results[[met]]$pairs_full %>%
      mutate(comparison = "family:treatment"),
    stats_results[[met]]$pairs_family %>%
      mutate(comparison = "family"),
    stats_results[[met]]$pairs_trt %>%
      mutate(comparison = "treatment")
  ) %>% mutate(metric = met)
})

write_csv(stats_compiled,
          file.path(out_dir, "pairwise_stats.csv"))

message("Output files written to: ", out_dir)
]]> 2026 resazurin Magallana gigas Crassostrea gigas Pacific oyster heat stress https://robertslab.github.io/sams-notebook/posts/2026/2026-07-13-Resazurin-Assays---USDA-M.gigas-Families-in-Response-to-Temperature-Stress/ Mon, 13 Jul 2026 07:00:00 GMT Resazurin Assays - USDA M.gigas Families in Response to Temperature Stress Sam White https://robertslab.github.io/sams-notebook/posts/2026/2026-06-24-Resazurin-Assays---USDA-M.gigas-Families-in-Response-to-Temperature-Stress/

INTRO

Notebook was knitted from 01.00-resazurin-20260624-mgig-33C.Rmd

RESULTS

Resazurin-based metabolic activity was measured in nine USDA M. gigas families (1, 2, 3, 5, 6, 7, 8, 9, 10) at four timepoints (0, 1, 2, 3 h) during 33°C heat stress, with fluorescence normalized to shell area (mm²) across three replicate measurement rounds (n = 18 individuals per family).

Metabolic activity increased significantly over time (Type III LME ANOVA: F₃,₄₅₉ = 389.9, p < 0.0001), confirming that the resazurin assay detected a biologically meaningful response to heat exposure. However, no significant differences in metabolism were detected among families, either as a main effect (F₈,₁₅₃ = 0.95, p = 0.48) or as a family × time interaction (F₂₄,₄₅₉ = 0.94, p = 0.54).

AUC (area under the curve) analysis corroborated these findings: the one-way ANOVA on per-individual AUC values showed no significant family effect (F₈,₁₅₃ = 1.03, p = 0.42), and no pairwise family comparisons reached significance after Tukey adjustment (all p ≥ 0.24). Family 3 had the numerically highest mean AUC (0.00377 ± 0.000414 SE) and Family 1 the lowest (0.00234 ± 0.000157 SE), but these differences were not statistically distinguishable given the high within-family variability.

1 Background

Adult oysters from nine USDA families were placed in plastic cups and submerged in ~100 mL of freshly prepared resazurin working solution at room temperature (~22°C). Cups were incubated in a water table at 33°C. Three rounds of measurements were performed, with different individuals from each family used in each round. At designated timepoints, 200 µL of resazurin was sampled from each cup and placed in a 96-well plate. Fluorescence was measured using a Synergy HTX (Agilent) plate reader.

See Resazurin/data/20260624-mgig-33C/README.md for full experimental notes.

1.1 Expected inputs

Path Description
Resazurin/data/20260624-mgig-33C/plate-*-T*.txt Plate reader fluorescence exports (one file per plate per timepoint)
Resazurin/data/20260624-mgig-33C/layout.csv Well metadata: plate ID, well ID, blank flag, family groups, cup ID (cup_ID.group), measurement round (round.group), unique individual ID (sample_ID.group), area measurements (mm², from ImageJ)

1.2 Expected outputs

All outputs are written to Resazurin/outputs/01.00-resazurin-20260624-mgig-33C/.

File Description
figures/ All plots generated by this script
auc_all_metrics.csv Per-individual AUC values for every active measurement metric
auc_summary.csv Group-level AUC summary statistics (mean, SD, SE, median)
metabolism.csv Full per-well per-timepoint metabolism data frame
pairwise_stats.csv Tukey-adjusted pairwise comparisons from AUC linear models

2 Setup

2.1 Knitr options

knitr::opts_chunk$set(
  echo = TRUE,         # Display code chunks
  eval = TRUE,        # Evaluate code chunks
  warning = FALSE,     # Hide warnings
  message = FALSE,     # Hide messages
  comment = "",         # Prevents appending '##' to beginning of lines in code output
  results = 'hold'     # Holds output so it's all printed together after code chunk
)

2.2 Load libraries

library(tidyverse)
library(pracma)       # trapz()
library(lme4)
library(lmerTest)
library(emmeans)
library(multcompView)
library(cowplot)
library(colorspace)   # qualitative_hcl() for large palettes

3 Helper Functions

normalize_well_id <- function(x) {
  x <- toupper(trimws(x))
  valid <- str_detect(x, "^[A-Z]+[0-9]+$")
  out <- rep(NA_character_, length(x))
  if (!any(valid)) return(out)
  m <- str_match(x[valid], "^([A-Z]+)([0-9]+)$")
  out[valid] <- paste0(m[, 2], as.integer(m[, 3]))
  out
}

parse_time_hr <- function(path) {
  hit <- str_match(basename(path),
                   "(?i)[_-]T([0-9]+(?:\\.[0-9]+)?)\\.txt$")
  as.numeric(hit[, 2])
}

parse_plate_id <- function(path) {
  hit <- str_match(basename(path),
    "(?i)^[0-9]{8}_(round[0-9]+)_T[0-9]+(?:\\.[0-9]+)?\\.txt$")
  id <- hit[, 2]
  ifelse(is.na(id), "unknown", id)
}

extract_results_block <- function(lines) {
  results_idx <- which(trimws(lines) == "Results")
  if (length(results_idx) == 0) stop("No Results section found")
  idx <- results_idx[1]
  header_tokens <- str_split(lines[idx + 1], "\\t")[[1]] |> trimws()
  col_ids <- header_tokens[
    header_tokens != "" & str_detect(header_tokens, "^[0-9]+$")]
  j <- idx + 2
  data_lines <- character()
  while (j <= length(lines)) {
    line <- lines[j]
    if (trimws(line) == "") break
    if (!str_detect(line, "^[A-Za-z]\\t")) break
    data_lines <- c(data_lines, line)
    j <- j + 1
  }
  list(col_ids = col_ids, data_lines = data_lines)
}

parse_plate_export <- function(path) {
  lines <- readLines(path, warn = FALSE)
  res <- extract_results_block(lines)

  map_dfr(res$data_lines, function(line) {
    tokens <- str_split(line, "\\t")[[1]] |> trimws()
    tokens <- tokens[tokens != ""]
    row_letter <- tokens[1]
    nums <- suppressWarnings(as.numeric(tokens[-1]))
    valid_idx <- which(!is.na(nums))
    if (length(valid_idx) == 0) return(tibble())
    vals <- nums[valid_idx]
    n <- min(length(vals), length(res$col_ids))
    tibble(
      row_id  = toupper(row_letter),
      col_id  = as.integer(res$col_ids[seq_len(n)]),
      well_id = normalize_well_id(
        paste0(toupper(row_letter), res$col_ids[seq_len(n)])),
      value   = vals[seq_len(n)]
    )
  }) %>%
    mutate(
      plate_id = str_to_lower(parse_plate_id(path)),
      time_hr  = parse_time_hr(path)
    )
}

trapezoid_auc <- function(time_hr, value) {
  ok <- is.finite(time_hr) & is.finite(value)
  t <- time_hr[ok]
  v <- value[ok]
  if (length(t) < 2) return(NA_real_)
  ord <- order(t)
  t <- t[ord]; v <- v[ord]
  sum(diff(t) * (head(v, -1) + tail(v, -1)) / 2)
}

# Shared helper: extract display unit string from a measurement column name.
# e.g. "area_mm2_measurement" -> "mm²", "weight_mg_measurement" -> "mg"
parse_meas_unit <- function(col_name) {
  unit_raw <- col_name |>
    str_remove("^metabolism_per_") |>
    str_remove("_measurement$") |>
    str_extract("[^_]+$")
  case_when(
    unit_raw == "mm2" ~ "mm²",
    unit_raw == "cm2" ~ "cm²",
    unit_raw == "mm3" ~ "mm³",
    unit_raw == "cm3" ~ "cm³",
    TRUE              ~ unit_raw
  )
}

# y-axis label for metabolism line plots: "fold change/mm²"
metabolism_y_label <- function(col_name) {
  paste0("Metabolism (fold change/", parse_meas_unit(col_name), ")")
}

# y-axis label for AUC box plots: "Metabolism (AUC; mm²)"
auc_y_label <- function(metric_name) {
  paste0("Metabolism (AUC; ", parse_meas_unit(metric_name), ")")
}

4 Load Data

4.1 Plate export files

proj_root <- rprojroot::find_rstudio_root_file()
data_dir  <- file.path(proj_root, "Resazurin", "data", "20260624-mgig-33C")
out_dir   <- file.path(proj_root, "Resazurin", "outputs",
                        "01.00-resazurin-20260624-mgig-33C")
fig_dir   <- file.path(out_dir, "figures")

dir.create(fig_dir, recursive = TRUE, showWarnings = FALSE)
dir.create(out_dir, recursive = TRUE, showWarnings = FALSE)

plate_files <- list.files(
  data_dir,
  pattern = "(?i)^[0-9]{8}_round[0-9]+_T[0-9]+(?:\\.[0-9]+)?\\.txt$",
  full.names = TRUE
)

plate_raw <- map_dfr(plate_files, function(path) {
  tryCatch(parse_plate_export(path),
           error = function(e) {
             message("Parse error in ", basename(path), ": ", e$message)
             tibble()
           })
})

str(plate_raw)
tibble [1,152 × 6] (S3: tbl_df/tbl/data.frame)
 $ row_id  : chr [1:1152] "A" "A" "A" "A" ...
 $ col_id  : int [1:1152] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id : chr [1:1152] "A1" "A2" "A3" "A4" ...
 $ value   : num [1:1152] 627 598 599 580 563 752 577 559 568 565 ...
 $ plate_id: chr [1:1152] "round1" "round1" "round1" "round1" ...
 $ time_hr : num [1:1152] 0 0 0 0 0 0 0 0 0 0 ...

4.2 Plate consistency check

Checks that every plate has the same number of wells at every timepoint. The expected well count is the mode across all plate × timepoint reads. Any plate with at least one deviating read is flagged and dropped entirely before any further analysis — removing only the aberrant timepoint would break the fold-change baseline calculation.

well_counts <- plate_raw %>%
  group_by(plate_id, time_hr) %>%
  summarise(n_wells = n_distinct(well_id), .groups = "drop")

expected_n_wells <- as.integer(
  names(which.max(table(well_counts$n_wells)))
)

inconsistent_reads <- well_counts %>%
  filter(n_wells != expected_n_wells) %>%
  arrange(plate_id, time_hr)

inconsistent_plate_ids <- unique(inconsistent_reads$plate_id)

if (nrow(inconsistent_reads) > 0) {
  cat("**Plate consistency check FAILED.**",
      "Expected", expected_n_wells, "wells per plate-timepoint read.",
      length(inconsistent_plate_ids),
      "plate(s) have at least one deviating read and are excluded",
      "from all analyses:\n\n")
  cat(knitr::kable(
    inconsistent_reads,
    col.names = c("Plate", "Time (h)", "Wells read"),
    caption   = paste("Expected:", expected_n_wells, "wells per read")
  ), sep = "\n")
  cat("\n")
  plate_raw <- plate_raw %>%
    filter(!plate_id %in% inconsistent_plate_ids)
  message(length(inconsistent_plate_ids),
          " plate(s) removed from plate_raw: ",
          paste(inconsistent_plate_ids, collapse = ", "))
} else {
  cat("Plate consistency check passed: all",
      n_distinct(well_counts$plate_id), "plates have",
      expected_n_wells, "wells at every timepoint.\n")
}

Plate consistency check passed: all 3 plates have 96 wells at every timepoint.

4.3 Layout file

layout_path <- file.path(data_dir, "layout.csv")

layout_raw <- read_csv(layout_path,
                       col_types = cols(.default = "c"),
                       show_col_types = FALSE)

# Standardise column names to snake_case
names(layout_raw) <- names(layout_raw) |>
  str_to_lower() |>
  str_replace_all("[^a-z0-9]+", "_") |>
  str_replace_all("_+", "_") |>
  str_replace("_$", "")

# Normalise plate_id to match plate file ids (strip "plate-" prefix)
layout_clean <- layout_raw %>%
  mutate(
    plate_id = str_remove(str_to_lower(plate_id), "^plate-"),
    well_id  = normalize_well_id(plate_well),
    is_blank = if ("is_blank" %in% names(layout_raw))
      toupper(trimws(is_blank)) %in% c("TRUE", "T", "1", "YES", "Y")
    else
      FALSE
  )

found_exclude_col <- intersect(
  c("exclude_from_analysis", "exclude", "omit", "not_analyzed"),
  names(layout_clean)
)[1]
layout_clean <- layout_clean %>%
  mutate(
    exclude_from_analysis = if (!is.na(found_exclude_col))
      toupper(trimws(.data[[found_exclude_col]])) %in%
        c("TRUE", "T", "1", "YES", "Y")
    else
      FALSE
  )

# Identify measurement columns and group columns
measurement_cols <- names(layout_clean)[
  str_detect(names(layout_clean), "_measurement$")]
group_cols <- names(layout_clean)[
  str_detect(names(layout_clean), "_group$")]

# Cast measurement columns to numeric
layout_clean <- layout_clean %>%
  mutate(across(all_of(measurement_cols),
                ~ suppressWarnings(as.numeric(.x))))

# Determine which measurement columns actually contain finite data
active_meas_cols <- measurement_cols[
  sapply(measurement_cols, function(col)
    any(is.finite(layout_clean[[col]]), na.rm = TRUE))]

# Normalise group values to lowercase so they match colour scale definitions
layout_clean <- layout_clean %>%
  mutate(across(all_of(group_cols),
                ~ str_to_lower(trimws(as.character(.x)))))

message("Group columns: ", paste(group_cols, collapse = ", "))
message("Active measurement columns: ",
        paste(active_meas_cols, collapse = ", "))

str(layout_clean)
tibble [180 × 16] (S3: tbl_df/tbl/data.frame)
 $ plate_id             : chr [1:180] "round1" "round1" "round1" "round1" ...
 $ plate_well           : chr [1:180] "A1" "A2" "A3" "A4" ...
 $ is_blank             : logi [1:180] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ family_id_group      : chr [1:180] "1" "2" "3" "5" ...
 $ cup_id_group         : chr [1:180] "1" "2" "3" "4" ...
 $ round_group          : chr [1:180] "1" "1" "1" "1" ...
 $ sample_id_group      : chr [1:180] "1_01" "1_02" "1_03" "1_04" ...
 $ exclude_from_analysis: logi [1:180] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:180] NA NA NA NA ...
 $ weight_g_measurement : num [1:180] NA NA NA NA NA NA NA NA NA NA ...
 $ width_mm_measurement : num [1:180] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:180] NA NA NA NA NA NA NA NA NA NA ...
 $ treatment_group      : chr [1:180] NA NA NA NA ...
 $ area_mm2_measurement : num [1:180] 1441 1594 1362 1860 1407 ...
 $ imagej_id            : chr [1:180] "2" "2" "1" "1" ...
 $ well_id              : chr [1:180] "A1" "A2" "A3" "A4" ...

5 Merge Plate Data with Layout

dat <- plate_raw %>%
  left_join(
    layout_clean %>%
      select(plate_id, well_id, is_blank, exclude_from_analysis,
             any_of("exclude_reason"),
             all_of(group_cols), all_of(measurement_cols)),
    by = c("plate_id", "well_id")
  ) %>%
  mutate(
    is_blank = replace_na(is_blank, FALSE),
    exclude_from_analysis = replace_na(exclude_from_analysis, FALSE)
  )

str(dat)
tibble [1,152 × 18] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:1152] "A" "A" "A" "A" ...
 $ col_id               : int [1:1152] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id              : chr [1:1152] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:1152] 627 598 599 580 563 752 577 559 568 565 ...
 $ plate_id             : chr [1:1152] "round1" "round1" "round1" "round1" ...
 $ time_hr              : num [1:1152] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:1152] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:1152] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:1152] NA NA NA NA ...
 $ family_id_group      : chr [1:1152] "1" "2" "3" "5" ...
 $ cup_id_group         : chr [1:1152] "1" "2" "3" "4" ...
 $ round_group          : chr [1:1152] "1" "1" "1" "1" ...
 $ sample_id_group      : chr [1:1152] "1_01" "1_02" "1_03" "1_04" ...
 $ treatment_group      : chr [1:1152] NA NA NA NA ...
 $ weight_g_measurement : num [1:1152] NA NA NA NA NA NA NA NA NA NA ...
 $ width_mm_measurement : num [1:1152] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:1152] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:1152] 1441 1594 1362 1860 1407 ...

6 Raw Fluorescence

6.1 Data frame

# Wells in the plate reader output that have no layout entry get all-NA group
# columns after the join. Keep only wells assigned to at least one group.
active_gc <- intersect(group_cols, names(dat))

raw_df <- dat %>%
  filter(
    !is_blank,
    if (length(active_gc) > 0)
      if_any(all_of(active_gc), ~ !is.na(.))
    else
      TRUE
  ) %>%
  mutate(
    trace_id = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  )

families   <- str_sort(unique(na.omit(raw_df$family_id_group)), numeric = TRUE)
treatments <- sort(unique(na.omit(raw_df$treatment_group)))

n_fam <- length(families)
n_trt <- length(treatments)

# Palette strategy:
#   <= 7 groups : Okabe-Ito (gold standard for colorblind-safe figures).
#   >  7 groups : colorspace::qualitative_hcl("Dynamic") scales to any N
#                 using perceptually uniform HCL space — no colour collisions.
# Black (#000000) is excluded from both and reserved for blank wells.
okabe_ito_7 <- c(
  "#E69F00", "#56B4E9", "#009E73", "#F0E442",
  "#0072B2", "#D55E00", "#CC79A7"
)
make_palette <- function(n) {
  if (n == 0L) return(character(0))
  if (n <= length(okabe_ito_7)) return(okabe_ito_7[seq_len(n)])
  colorspace::qualitative_hcl(n, palette = "Dynamic")
}

all_colours   <- make_palette(n_fam + n_trt)
fam_colours   <- setNames(all_colours[seq_len(n_fam)], families)
trt_colours   <- setNames(all_colours[n_fam + seq_len(n_trt)], treatments)

lty_pool <- c("solid", "dashed", "dotted", "dotdash", "longdash")
trt_linetypes <- setNames(
  lty_pool[(seq_len(n_trt) - 1L) %% length(lty_pool) + 1L],
  treatments
)
plate_well_colours <- c(blank = "black", fam_colours)

has_trt <- n_trt > 0

str(raw_df)
tibble [648 × 19] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:648] "A" "A" "A" "A" ...
 $ col_id               : int [1:648] 1 2 3 4 5 6 7 8 9 11 ...
 $ well_id              : chr [1:648] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:648] 627 598 599 580 563 752 577 559 568 596 ...
 $ plate_id             : chr [1:648] "round1" "round1" "round1" "round1" ...
 $ time_hr              : num [1:648] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:648] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:648] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:648] NA NA NA NA ...
 $ family_id_group      : chr [1:648] "1" "2" "3" "5" ...
 $ cup_id_group         : chr [1:648] "1" "2" "3" "4" ...
 $ round_group          : chr [1:648] "1" "1" "1" "1" ...
 $ sample_id_group      : chr [1:648] "1_01" "1_02" "1_03" "1_04" ...
 $ treatment_group      : chr [1:648] NA NA NA NA ...
 $ weight_g_measurement : num [1:648] NA NA NA NA NA NA NA NA NA NA ...
 $ width_mm_measurement : num [1:648] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:648] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:648] 1441 1594 1362 1860 1407 ...
 $ trace_id             : chr [1:648] "1_01" "1_02" "1_03" "1_04" ...

6.2 Raw fluorescence by plate (including blanks)

p_raw_plates <- dat %>%
  filter(is.finite(time_hr), is.finite(value)) %>%
  mutate(
    colour_group = if_else(is_blank, "blank",
                           coalesce(family_id_group, "sample")),
    trace_id     = paste(plate_id, well_id, sep = "_")
  ) %>%
  ggplot(aes(x = time_hr, y = value,
             group = trace_id, colour = colour_group)) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1, alpha = 0.7) +
  facet_wrap(~ plate_id) +
  scale_colour_manual(
    values   = plate_well_colours,
    name     = "Group",
    breaks   = names(plate_well_colours),
    na.value = "grey80"
  ) +
  labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_raw_plates

ggsave(file.path(fig_dir, "raw_fluor_by_plate.png"),
       p_raw_plates, width = 10, height = 8)

6.3 Mean raw fluorescence by family

raw_family_summary <- raw_df %>%
  filter(!is.na(family_id_group), !exclude_from_analysis) %>%
  group_by(family_id_group, treatment_group, time_hr) %>%
  summarise(
    mean_fluor = mean(value, na.rm = TRUE),
    se_fluor   = sd(value, na.rm = TRUE) /
      sqrt(sum(!is.na(value))),
    n          = sum(!is.na(value)),
    .groups    = "drop"
  ) %>%
  mutate(group_var = if (has_trt)
    paste(family_id_group, treatment_group, sep = ".")
  else
    family_id_group)

p_raw_mean <- ggplot(raw_family_summary,
    aes(x = time_hr, y = mean_fluor,
        colour = family_id_group,
        group = group_var)) +
  geom_ribbon(aes(ymin = mean_fluor - se_fluor,
                  ymax = mean_fluor + se_fluor,
                  fill = family_id_group),
              alpha = 0.15, colour = NA) +
  geom_line(
    mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
    linewidth = 1) +
  geom_point(size = 2) +
  scale_colour_manual(values = fam_colours, name = "Family",
                      breaks = families) +
  scale_fill_manual(values = fam_colours, name = "Family",
                    breaks = families) +
  labs(x = "Time (h)", y = "Mean raw fluorescence (RFU ± SE)") +
  theme_classic(base_size = 13) +
  if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

p_raw_mean

ggsave(file.path(fig_dir, "raw_mean_by_family.png"),
       p_raw_mean, width = 8, height = 5)

6.4 Individual raw fluorescence traces by family

p_raw_by_family <- raw_df %>%
  filter(!is.na(family_id_group)) %>%
  ggplot(aes(x = time_hr, y = value, group = trace_id,
             colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1.2, alpha = 0.7) +
  facet_wrap(~ family_id_group) +
  scale_colour_manual(
    values = if (has_trt) trt_colours else fam_colours,
    name   = if (has_trt) "Treatment" else "Family",
    breaks = if (has_trt) treatments else families) +
  labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_raw_by_family

ggsave(file.path(fig_dir, "raw_individual_by_family.png"),
       p_raw_by_family, width = 10, height = 5)

6.5 Individual raw fluorescence traces by treatment

if (has_trt) {
  p_raw_by_treatment <- raw_df %>%
    ggplot(aes(x = time_hr, y = value,
               group = trace_id, colour = family_id_group)) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ treatment_group) +
    scale_colour_manual(values = fam_colours, name = "Family",
                        breaks = families) +
    labs(x = "Time (h)", y = "Raw fluorescence (RFU)") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_raw_by_treatment
  ggsave(file.path(fig_dir, "raw_individual_by_treatment.png"),
         p_raw_by_treatment, width = 10, height = 5)
}

6.6 Excluded samples

Wells flagged exclude_from_analysis = TRUE appear in the raw fluorescence plots above but are omitted from all analyses that follow.

excluded_wells <- dat %>%
  filter(!is_blank, exclude_from_analysis) %>%
  mutate(
    sample = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  ) %>%
  select(plate_id, well_id, sample, family_id_group, treatment_group,
         any_of("exclude_reason")) %>%
  distinct() %>%
  arrange(plate_id, well_id)

if (nrow(excluded_wells) > 0) {
  col_names <- c("Plate", "Well", "Sample", "Family", "Treatment")
  if ("exclude_reason" %in% names(excluded_wells))
    col_names <- c(col_names, "Reason")
  cat(knitr::kable(excluded_wells, col.names = col_names), sep = "\n")
} else {
  cat("No wells are excluded from analysis.\n")
}

No wells are excluded from analysis.

7 Blank Correction via Fold-Change Normalization

T0 is the earliest timepoint present in the dataset (not necessarily 0 hr). Sample fold-change is expressed relative to each individual’s T0 reading, resolved by sample_id_group when that column is populated — allowing the same animal to be tracked across plates — or by plate_id + well_id when no sample IDs exist (backward-compatible with single-plate, multi-timepoint designs). Blank fold-change is the per-plate mean blank RFU at each timepoint divided by the pooled mean blank RFU at T0. Subtracting blank fold-change from sample fold-change removes background fluorescence drift; all samples start at exactly 0 at T0 by construction.

7.1 Step 1 – Identify T0 and compute per-sample fold-change

# T0 = earliest timepoint present in the dataset
t0_time <- min(dat$time_hr[is.finite(dat$time_hr)], na.rm = TRUE)
message("T0 timepoint: ", t0_time, " hr")

# T0 reference value per individual.
# Resolved by sample_id_group (cross-plate tracking) when available;
# falls back to plate+well for layouts without explicit sample IDs.
t0_all <- dat %>%
  filter(time_hr == t0_time, !is_blank, is.finite(value)) %>%
  mutate(sample_key = if_else(
    !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
    as.character(sample_id_group),
    paste(plate_id, well_id, sep = "_")
  )) %>%
  group_by(sample_key) %>%
  summarise(value_t0 = mean(value, na.rm = TRUE), .groups = "drop")

dat_fc <- dat %>%
  mutate(sample_key = if_else(
    !is_blank &
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
    as.character(sample_id_group),
    paste(plate_id, well_id, sep = "_")
  )) %>%
  left_join(t0_all, by = "sample_key") %>%
  mutate(fold_change = if_else(
    !is_blank & is.finite(value_t0) & value_t0 > 0,
    value / value_t0,
    NA_real_
  ))

str(dat_fc)
tibble [1,152 × 21] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:1152] "A" "A" "A" "A" ...
 $ col_id               : int [1:1152] 1 2 3 4 5 6 7 8 9 10 ...
 $ well_id              : chr [1:1152] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:1152] 627 598 599 580 563 752 577 559 568 565 ...
 $ plate_id             : chr [1:1152] "round1" "round1" "round1" "round1" ...
 $ time_hr              : num [1:1152] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:1152] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:1152] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:1152] NA NA NA NA ...
 $ family_id_group      : chr [1:1152] "1" "2" "3" "5" ...
 $ cup_id_group         : chr [1:1152] "1" "2" "3" "4" ...
 $ round_group          : chr [1:1152] "1" "1" "1" "1" ...
 $ sample_id_group      : chr [1:1152] "1_01" "1_02" "1_03" "1_04" ...
 $ treatment_group      : chr [1:1152] NA NA NA NA ...
 $ weight_g_measurement : num [1:1152] NA NA NA NA NA NA NA NA NA NA ...
 $ width_mm_measurement : num [1:1152] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:1152] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:1152] 1441 1594 1362 1860 1407 ...
 $ sample_key           : chr [1:1152] "1_01" "1_02" "1_03" "1_04" ...
 $ value_t0             : num [1:1152] 627 598 599 580 563 752 577 559 568 NA ...
 $ fold_change          : num [1:1152] 1 1 1 1 1 1 1 1 1 NA ...

7.2 Step 2 – Blank fold-change reference per plate per timepoint

# Pooled mean blank RFU at T0 across all T0 plates
mean_blank_t0 <- dat %>%
  filter(is_blank, time_hr == t0_time, is.finite(value)) %>%
  pull(value) %>%
  mean(na.rm = TRUE)

if (!is.finite(mean_blank_t0))
  message("No blank readings found at T0 (", t0_time,
          " hr); blank correction will produce NA.")

# Per-plate per-timepoint mean blank expressed as fold-change relative to T0
blank_fc_ref <- dat %>%
  filter(is_blank, is.finite(value)) %>%
  group_by(plate_id, time_hr) %>%
  summarise(mean_blank_rfu = mean(value, na.rm = TRUE), .groups = "drop") %>%
  mutate(mean_blank_fc = mean_blank_rfu / mean_blank_t0)

str(blank_fc_ref)
tibble [12 × 4] (S3: tbl_df/tbl/data.frame)
 $ plate_id      : chr [1:12] "round1" "round1" "round1" "round1" ...
 $ time_hr       : num [1:12] 0 1 2 3 0 1 2 3 0 1 ...
 $ mean_blank_rfu: num [1:12] 568 715 612 630 632 ...
 $ mean_blank_fc : num [1:12] 0.979 1.232 1.055 1.085 1.089 ...

7.3 Step 3 – Subtract blank fold-change from sample fold-change

samples <- dat_fc %>%
  filter(!is_blank, !exclude_from_analysis) %>%
  mutate(
    trace_id = if_else(
      !is.na(sample_id_group) & trimws(as.character(sample_id_group)) != "",
      as.character(sample_id_group),
      paste(plate_id, well_id, sep = "_")
    )
  ) %>%
  left_join(blank_fc_ref, by = c("plate_id", "time_hr")) %>%
  mutate(corrected_fc = fold_change - mean_blank_fc)

str(samples)
tibble [1,080 × 25] (S3: tbl_df/tbl/data.frame)
 $ row_id               : chr [1:1080] "A" "A" "A" "A" ...
 $ col_id               : int [1:1080] 1 2 3 4 5 6 7 8 9 11 ...
 $ well_id              : chr [1:1080] "A1" "A2" "A3" "A4" ...
 $ value                : num [1:1080] 627 598 599 580 563 752 577 559 568 596 ...
 $ plate_id             : chr [1:1080] "round1" "round1" "round1" "round1" ...
 $ time_hr              : num [1:1080] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank             : logi [1:1080] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis: logi [1:1080] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason       : chr [1:1080] NA NA NA NA ...
 $ family_id_group      : chr [1:1080] "1" "2" "3" "5" ...
 $ cup_id_group         : chr [1:1080] "1" "2" "3" "4" ...
 $ round_group          : chr [1:1080] "1" "1" "1" "1" ...
 $ sample_id_group      : chr [1:1080] "1_01" "1_02" "1_03" "1_04" ...
 $ treatment_group      : chr [1:1080] NA NA NA NA ...
 $ weight_g_measurement : num [1:1080] NA NA NA NA NA NA NA NA NA NA ...
 $ width_mm_measurement : num [1:1080] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement: num [1:1080] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement : num [1:1080] 1441 1594 1362 1860 1407 ...
 $ sample_key           : chr [1:1080] "1_01" "1_02" "1_03" "1_04" ...
 $ value_t0             : num [1:1080] 627 598 599 580 563 752 577 559 568 596 ...
 $ fold_change          : num [1:1080] 1 1 1 1 1 1 1 1 1 1 ...
 $ trace_id             : chr [1:1080] "1_01" "1_02" "1_03" "1_04" ...
 $ mean_blank_rfu       : num [1:1080] 568 568 568 568 568 ...
 $ mean_blank_fc        : num [1:1080] 0.979 0.979 0.979 0.979 0.979 ...
 $ corrected_fc         : num [1:1080] 0.0206 0.0206 0.0206 0.0206 0.0206 ...

8 Blank-Corrected Fold-Change

8.1 Mean by family

bc_fc_summary <- samples %>%
  filter(!is.na(family_id_group), !exclude_from_analysis) %>%
  group_by(family_id_group, treatment_group, time_hr) %>%
  summarise(
    mean_val = mean(corrected_fc, na.rm = TRUE),
    se_val   = sd(corrected_fc, na.rm = TRUE) /
      sqrt(sum(!is.na(corrected_fc))),
    n        = sum(!is.na(corrected_fc)),
    .groups  = "drop"
  ) %>%
  mutate(group_var = if (has_trt)
    paste(family_id_group, treatment_group, sep = ".")
  else
    family_id_group)

p_bc_fc_mean <- ggplot(bc_fc_summary,
    aes(x = time_hr, y = mean_val,
        colour = family_id_group,
        group = group_var)) +
  geom_ribbon(aes(ymin = mean_val - se_val,
                  ymax = mean_val + se_val,
                  fill = family_id_group),
              alpha = 0.15, colour = NA) +
  geom_line(
    mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
    linewidth = 1) +
  geom_point(size = 2) +
  scale_colour_manual(values = fam_colours, name = "Family",
                      breaks = families) +
  scale_fill_manual(values = fam_colours, name = "Family",
                    breaks = families) +
  labs(x = "Time (h)",
       y = "Mean blank-corrected fold-change (± SE)") +
  theme_classic(base_size = 13) +
  if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

p_bc_fc_mean

ggsave(file.path(fig_dir, "blank_corrected_fc_mean_by_family.png"),
       p_bc_fc_mean, width = 8, height = 5)

8.2 Individual traces by family

p_bc_fc_by_family <- samples %>%
  filter(!is.na(family_id_group)) %>%
  ggplot(aes(x = time_hr, y = corrected_fc, group = trace_id,
             colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
  geom_line(alpha = 0.6) +
  geom_point(size = 1.2, alpha = 0.7) +
  facet_wrap(~ family_id_group) +
  scale_colour_manual(
    values = if (has_trt) trt_colours else fam_colours,
    name   = if (has_trt) "Treatment" else "Family",
    breaks = if (has_trt) treatments else families) +
  labs(x = "Time (h)", y = "Blank-corrected fold-change") +
  theme_classic(base_size = 12) +
  theme(strip.background = element_blank(),
        strip.text       = element_text(face = "bold"))

p_bc_fc_by_family

ggsave(file.path(fig_dir, "blank_corrected_fc_by_family.png"),
       p_bc_fc_by_family, width = 10, height = 5)

8.3 Individual blank-corrected fold-change traces by treatment

if (has_trt) {
  p_bc_fc_by_treatment <- samples %>%
    ggplot(aes(x = time_hr, y = corrected_fc,
               group = trace_id, colour = family_id_group)) +
    geom_line(alpha = 0.6) +
    geom_point(size = 1.2, alpha = 0.7) +
    facet_wrap(~ treatment_group) +
    scale_colour_manual(values = fam_colours, name = "Family",
                        breaks = families) +
    labs(x = "Time (h)", y = "Blank-corrected fold-change") +
    theme_classic(base_size = 12) +
    theme(strip.background = element_blank(),
          strip.text       = element_text(face = "bold"))

  p_bc_fc_by_treatment
  ggsave(file.path(fig_dir, "blank_corrected_fc_by_treatment.png"),
         p_bc_fc_by_treatment, width = 10, height = 5)
}

9 Metabolism (Size-Normalised Fold-Change)

Blank-corrected fold-change divided by each active measurement column. This is “metabolism” as defined in Huffmyer et al.

if (length(active_meas_cols) == 0) {
  message("No active measurement columns: skipping metabolism calculation.")
  metabolism_df <- tibble()
} else {
  metabolism_df <- samples
  for (mc in active_meas_cols) {
    out_col <- paste0("metabolism_per_", mc)
    metabolism_df <- metabolism_df %>%
      mutate(!!out_col := if_else(
        is.finite(.data[[mc]]) & .data[[mc]] > 0 &
          is.finite(corrected_fc),
        corrected_fc / .data[[mc]],
        NA_real_
      ))
  }
}

str(metabolism_df)
tibble [1,080 × 26] (S3: tbl_df/tbl/data.frame)
 $ row_id                             : chr [1:1080] "A" "A" "A" "A" ...
 $ col_id                             : int [1:1080] 1 2 3 4 5 6 7 8 9 11 ...
 $ well_id                            : chr [1:1080] "A1" "A2" "A3" "A4" ...
 $ value                              : num [1:1080] 627 598 599 580 563 752 577 559 568 596 ...
 $ plate_id                           : chr [1:1080] "round1" "round1" "round1" "round1" ...
 $ time_hr                            : num [1:1080] 0 0 0 0 0 0 0 0 0 0 ...
 $ is_blank                           : logi [1:1080] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_from_analysis              : logi [1:1080] FALSE FALSE FALSE FALSE FALSE FALSE ...
 $ exclude_reason                     : chr [1:1080] NA NA NA NA ...
 $ family_id_group                    : chr [1:1080] "1" "2" "3" "5" ...
 $ cup_id_group                       : chr [1:1080] "1" "2" "3" "4" ...
 $ round_group                        : chr [1:1080] "1" "1" "1" "1" ...
 $ sample_id_group                    : chr [1:1080] "1_01" "1_02" "1_03" "1_04" ...
 $ treatment_group                    : chr [1:1080] NA NA NA NA ...
 $ weight_g_measurement               : num [1:1080] NA NA NA NA NA NA NA NA NA NA ...
 $ width_mm_measurement               : num [1:1080] NA NA NA NA NA NA NA NA NA NA ...
 $ length_mm_measurement              : num [1:1080] NA NA NA NA NA NA NA NA NA NA ...
 $ area_mm2_measurement               : num [1:1080] 1441 1594 1362 1860 1407 ...
 $ sample_key                         : chr [1:1080] "1_01" "1_02" "1_03" "1_04" ...
 $ value_t0                           : num [1:1080] 627 598 599 580 563 752 577 559 568 596 ...
 $ fold_change                        : num [1:1080] 1 1 1 1 1 1 1 1 1 1 ...
 $ trace_id                           : chr [1:1080] "1_01" "1_02" "1_03" "1_04" ...
 $ mean_blank_rfu                     : num [1:1080] 568 568 568 568 568 ...
 $ mean_blank_fc                      : num [1:1080] 0.979 0.979 0.979 0.979 0.979 ...
 $ corrected_fc                       : num [1:1080] 0.0206 0.0206 0.0206 0.0206 0.0206 ...
 $ metabolism_per_area_mm2_measurement: num [1:1080] 1.43e-05 1.29e-05 1.51e-05 1.11e-05 1.46e-05 ...

9.1 Mean metabolism by family

if (nrow(metabolism_df) > 0) {

  metab_cols <- paste0("metabolism_per_", active_meas_cols)

  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    metab_summary <- metabolism_df %>%
      filter(!is.na(family_id_group), !exclude_from_analysis) %>%
      group_by(family_id_group, treatment_group, time_hr) %>%
      summarise(
        mean_val = mean(.data[[col]], na.rm = TRUE),
        se_val   = sd(.data[[col]], na.rm = TRUE) /
          sqrt(sum(!is.na(.data[[col]]))),
        .groups  = "drop"
      ) %>%
      mutate(group_var = if (has_trt)
        paste(family_id_group, treatment_group, sep = ".")
      else
        family_id_group)

    p_metab_mean <- ggplot(metab_summary,
        aes(x = time_hr, y = mean_val,
            colour = family_id_group,
            group = group_var)) +
      geom_ribbon(aes(ymin = mean_val - se_val,
                      ymax = mean_val + se_val,
                      fill = family_id_group),
                  alpha = 0.15, colour = NA) +
      geom_line(
        mapping   = if (has_trt) aes(linetype = treatment_group) else NULL,
        linewidth = 1) +
      geom_point(size = 2) +
      scale_colour_manual(values = fam_colours, name = "Family",
                          breaks = families) +
      scale_fill_manual(values = fam_colours, name = "Family",
                        breaks = families) +
      labs(x = "Time (h)",
           y = paste0(metabolism_y_label(col), " (± SE)")) +
      theme_classic(base_size = 13) +
      if (has_trt) scale_linetype_manual(values = trt_linetypes, name = "Treatment") else NULL

    print(p_metab_mean)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_mean_", mc_label, ".png")),
      p_metab_mean, width = 8, height = 5)
  }
}

9.2 Individual metabolism traces by family

if (nrow(metabolism_df) > 0) {

  for (col in metab_cols) {
    if (!col %in% names(metabolism_df)) next
    mc_label <- str_remove(col, "^metabolism_per_")

    p_metab_by_family <- metabolism_df %>%
      filter(!is.na(family_id_group)) %>%
      ggplot(aes(x = time_hr, y = .data[[col]], group = trace_id,
                 colour = .data[[if (has_trt) "treatment_group" else "family_id_group"]])) +
      geom_line(alpha = 0.6) +
      geom_point(size = 1.2, alpha = 0.7) +
      facet_wrap(~ family_id_group) +
      scale_colour_manual(
        values = if (has_trt) trt_colours else fam_colours,
        name   = if (has_trt) "Treatment" else "Family",
        breaks = if (has_trt) treatments else families) +
      labs(x = "Time (h)", y = metabolism_y_label(col)) +
      theme_classic(base_size = 12) +
      theme(strip.background = element_blank(),
            strip.text       = element_text(face = "bold"))

    print(p_metab_by_family)
    ggsave(
      file.path(fig_dir,
                paste0("metabolism_individual_", mc_label, "_by_family.png")),
      p_metab_by_family, width = 10, height = 5)

    if (has_trt) {
      p_metab_by_treatment <- ggplot(metabolism_df,
          aes(x = time_hr, y = .data[[col]],
              group = trace_id, colour = family_id_group)) +
        geom_line(alpha = 0.6) +
        geom_point(size = 1.2, alpha = 0.7) +
        facet_wrap(~ treatment_group) +
        scale_colour_manual(values = fam_colours, name = "Family",
                            breaks = families) +
        labs(x = "Time (h)", y = metabolism_y_label(col)) +
        theme_classic(base_size = 12) +
        theme(strip.background = element_blank(),
              strip.text       = element_text(face = "bold"))

      print(p_metab_by_treatment)
      ggsave(
        file.path(fig_dir,
                  paste0("metabolism_individual_", mc_label, "_by_treatment.png")),
        p_metab_by_treatment, width = 10, height = 5)
    }
  }
}

10 Time-Series Statistical Analysis

Linear mixed effects models test the effect of experimental variables on metabolism over time. Individual (sample_id_group) is included as a random intercept to account for repeated measures across timepoints. Type III ANOVA with Satterthwaite’s approximation (lmerTest) assesses significance; post-hoc pairwise comparisons use estimated marginal means (emmeans, Tukey adjustment).

run_ts_stats <- function(df, value_col) {
  has_family    <- "family_id_group" %in% names(df) &&
    length(unique(na.omit(df$family_id_group))) > 1
  has_treatment <- "treatment_group" %in% names(df) &&
    length(unique(na.omit(df$treatment_group))) > 1

  if (!has_family && !has_treatment) return(NULL)

  df <- df %>%
    filter(is.finite(.data[[value_col]]), is.finite(time_hr)) %>%
    mutate(
      time_f     = factor(time_hr),
      individual = factor(trace_id)
    )

  if (nrow(df) == 0) return(NULL)

  if (has_family)    df <- df %>% mutate(family    = factor(family_id_group))
  if (has_treatment) df <- df %>% mutate(treatment = factor(treatment_group))

  if (has_family    && length(unique(na.omit(df$family)))    < 2) return(NULL)
  if (has_treatment && length(unique(na.omit(df$treatment))) < 2) return(NULL)

  fixed <- if (has_family && has_treatment) {
    paste0(value_col, " ~ time_f * family * treatment")
  } else if (has_family) {
    paste0(value_col, " ~ time_f * family")
  } else {
    paste0(value_col, " ~ time_f * treatment")
  }

  model <- lmer(
    as.formula(paste0(fixed, " + (1 | individual)")),
    data = df
  )

  anova_res <- anova(model, type = 3, ddf = "Satterthwaite")

  # Pairwise comparisons of group combinations at each timepoint
  emm_spec <- if (has_family && has_treatment) {
    ~ family * treatment | time_f
  } else if (has_family) {
    ~ family | time_f
  } else {
    ~ treatment | time_f
  }

  emm       <- emmeans(model, emm_spec)
  pairs_res <- as.data.frame(pairs(emm, adjust = "tukey"))

  # Main-effect marginal means (collapsed across time)
  emm_main <- if (has_family && has_treatment) {
    emmeans(model, ~ family * treatment)
  } else if (has_family) {
    emmeans(model, ~ family)
  } else {
    emmeans(model, ~ treatment)
  }

  pairs_main <- as.data.frame(pairs(emm_main, adjust = "tukey"))

  list(
    model         = model,
    anova         = anova_res,
    pairs_by_time = pairs_res,
    pairs_main    = pairs_main,
    has_family    = has_family,
    has_treatment = has_treatment
  )
}

ts_stats <- list()
if (nrow(metabolism_df) > 0) {
  for (mc in active_meas_cols) {
    col <- paste0("metabolism_per_", mc)
    if (col %in% names(metabolism_df))
      ts_stats[[col]] <- run_ts_stats(metabolism_df, col)
  }
}

10.1 Results

for (col in names(ts_stats)) {
  res <- ts_stats[[col]]
  if (is.null(res)) next

  cat("\n\n### Metric:", col, "\n\n")

  cat("**Type III ANOVA (Satterthwaite approximation):**\n\n")
  cat(knitr::kable(as.data.frame(res$anova), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")

  cat("**Marginal means – main effects (collapsed across time):**\n\n")
  cat(knitr::kable(as.data.frame(res$pairs_main), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")

  cat("**Pairwise comparisons by timepoint (Tukey):**\n\n")
  cat(knitr::kable(as.data.frame(res$pairs_by_time), digits = 4, format = "pipe"), sep = "\n")
  cat("\n")
}

10.1.1 Metric: metabolism_per_area_mm2_measurement

Type III ANOVA (Satterthwaite approximation):

Sum Sq Mean Sq NumDF DenDF F value Pr(>F)
time_f 3e-04 1e-04 3 459 389.9429 0.0000
family 0e+00 0e+00 8 153 0.9504 0.4772
time_f:family 0e+00 0e+00 24 459 0.9440 0.5411

Marginal means – main effects (collapsed across time):

contrast estimate SE df t.ratio p.value
family1 - family10 -1e-04 2e-04 153 -0.6072 0.9995
family1 - family2 -3e-04 2e-04 153 -1.6690 0.7643
family1 - family3 -4e-04 2e-04 153 -2.3926 0.2956
family1 - family5 -3e-04 2e-04 153 -1.5462 0.8316
family1 - family6 -3e-04 2e-04 153 -1.5621 0.8236
family1 - family7 -3e-04 2e-04 153 -1.6770 0.7595
family1 - family8 -2e-04 2e-04 153 -1.1174 0.9708
family1 - family9 -2e-04 2e-04 153 -1.2970 0.9309
family10 - family2 -2e-04 2e-04 153 -1.0618 0.9787
family10 - family3 -3e-04 2e-04 153 -1.7854 0.6917
family10 - family5 -2e-04 2e-04 153 -0.9390 0.9903
family10 - family6 -2e-04 2e-04 153 -0.9549 0.9892
family10 - family7 -2e-04 2e-04 153 -1.0698 0.9776
family10 - family8 -1e-04 2e-04 153 -0.5102 0.9999
family10 - family9 -1e-04 2e-04 153 -0.6898 0.9989
family2 - family3 -1e-04 2e-04 153 -0.7236 0.9984
family2 - family5 0e+00 2e-04 153 0.1227 1.0000
family2 - family6 0e+00 2e-04 153 0.1068 1.0000
family2 - family7 0e+00 2e-04 153 -0.0081 1.0000
family2 - family8 1e-04 2e-04 153 0.5516 0.9998
family2 - family9 1e-04 2e-04 153 0.3719 1.0000
family3 - family5 2e-04 2e-04 153 0.8463 0.9952
family3 - family6 2e-04 2e-04 153 0.8304 0.9958
family3 - family7 1e-04 2e-04 153 0.7155 0.9985
family3 - family8 2e-04 2e-04 153 1.2752 0.9371
family3 - family9 2e-04 2e-04 153 1.0955 0.9741
family5 - family6 0e+00 2e-04 153 -0.0159 1.0000
family5 - family7 0e+00 2e-04 153 -0.1308 1.0000
family5 - family8 1e-04 2e-04 153 0.4288 1.0000
family5 - family9 0e+00 2e-04 153 0.2492 1.0000
family6 - family7 0e+00 2e-04 153 -0.1149 1.0000
family6 - family8 1e-04 2e-04 153 0.4447 1.0000
family6 - family9 0e+00 2e-04 153 0.2651 1.0000
family7 - family8 1e-04 2e-04 153 0.5596 0.9998
family7 - family9 1e-04 2e-04 153 0.3800 1.0000
family8 - family9 0e+00 2e-04 153 -0.1796 1.0000

Pairwise comparisons by timepoint (Tukey):

contrast time_f estimate SE df t.ratio p.value
family1 - family10 0 0e+00 2e-04 340.5445 0.0072 1.0000
family1 - family2 0 0e+00 2e-04 340.5445 -0.0015 1.0000
family1 - family3 0 0e+00 2e-04 340.5445 -0.0038 1.0000
family1 - family5 0 0e+00 2e-04 340.5445 -0.0159 1.0000
family1 - family6 0 0e+00 2e-04 340.5445 -0.0190 1.0000
family1 - family7 0 0e+00 2e-04 340.5445 -0.0231 1.0000
family1 - family8 0 0e+00 2e-04 340.5445 0.0169 1.0000
family1 - family9 0 0e+00 2e-04 340.5445 -0.0200 1.0000
family10 - family2 0 0e+00 2e-04 340.5445 -0.0086 1.0000
family10 - family3 0 0e+00 2e-04 340.5445 -0.0109 1.0000
family10 - family5 0 0e+00 2e-04 340.5445 -0.0230 1.0000
family10 - family6 0 0e+00 2e-04 340.5445 -0.0261 1.0000
family10 - family7 0 0e+00 2e-04 340.5445 -0.0303 1.0000
family10 - family8 0 0e+00 2e-04 340.5445 0.0098 1.0000
family10 - family9 0 0e+00 2e-04 340.5445 -0.0271 1.0000
family2 - family3 0 0e+00 2e-04 340.5445 -0.0023 1.0000
family2 - family5 0 0e+00 2e-04 340.5445 -0.0144 1.0000
family2 - family6 0 0e+00 2e-04 340.5445 -0.0175 1.0000
family2 - family7 0 0e+00 2e-04 340.5445 -0.0216 1.0000
family2 - family8 0 0e+00 2e-04 340.5445 0.0184 1.0000
family2 - family9 0 0e+00 2e-04 340.5445 -0.0185 1.0000
family3 - family5 0 0e+00 2e-04 340.5445 -0.0121 1.0000
family3 - family6 0 0e+00 2e-04 340.5445 -0.0152 1.0000
family3 - family7 0 0e+00 2e-04 340.5445 -0.0194 1.0000
family3 - family8 0 0e+00 2e-04 340.5445 0.0207 1.0000
family3 - family9 0 0e+00 2e-04 340.5445 -0.0162 1.0000
family5 - family6 0 0e+00 2e-04 340.5445 -0.0031 1.0000
family5 - family7 0 0e+00 2e-04 340.5445 -0.0073 1.0000
family5 - family8 0 0e+00 2e-04 340.5445 0.0328 1.0000
family5 - family9 0 0e+00 2e-04 340.5445 -0.0041 1.0000
family6 - family7 0 0e+00 2e-04 340.5445 -0.0042 1.0000
family6 - family8 0 0e+00 2e-04 340.5445 0.0359 1.0000
family6 - family9 0 0e+00 2e-04 340.5445 -0.0010 1.0000
family7 - family8 0 0e+00 2e-04 340.5445 0.0401 1.0000
family7 - family9 0 0e+00 2e-04 340.5445 0.0031 1.0000
family8 - family9 0 0e+00 2e-04 340.5445 -0.0369 1.0000
family1 - family10 1 0e+00 2e-04 340.5445 -0.1282 1.0000
family1 - family2 1 -4e-04 2e-04 340.5445 -1.6047 0.8019
family1 - family3 1 -5e-04 2e-04 340.5445 -2.2758 0.3598
family1 - family5 1 -3e-04 2e-04 340.5445 -1.3509 0.9149
family1 - family6 1 -2e-04 2e-04 340.5445 -0.7646 0.9977
family1 - family7 1 -3e-04 2e-04 340.5445 -1.2464 0.9454
family1 - family8 1 -2e-04 2e-04 340.5445 -0.6868 0.9989
family1 - family9 1 -2e-04 2e-04 340.5445 -0.6725 0.9991
family10 - family2 1 -3e-04 2e-04 340.5445 -1.4765 0.8658
family10 - family3 1 -5e-04 2e-04 340.5445 -2.1476 0.4427
family10 - family5 1 -3e-04 2e-04 340.5445 -1.2227 0.9511
family10 - family6 1 -1e-04 2e-04 340.5445 -0.6363 0.9994
family10 - family7 1 -3e-04 2e-04 340.5445 -1.1181 0.9712
family10 - family8 1 -1e-04 2e-04 340.5445 -0.5585 0.9998
family10 - family9 1 -1e-04 2e-04 340.5445 -0.5442 0.9998
family2 - family3 1 -2e-04 2e-04 340.5445 -0.6711 0.9991
family2 - family5 1 1e-04 2e-04 340.5445 0.2538 1.0000
family2 - family6 1 2e-04 2e-04 340.5445 0.8402 0.9955
family2 - family7 1 1e-04 2e-04 340.5445 0.3583 1.0000
family2 - family8 1 2e-04 2e-04 340.5445 0.9180 0.9919
family2 - family9 1 2e-04 2e-04 340.5445 0.9323 0.9910
family3 - family5 1 2e-04 2e-04 340.5445 0.9249 0.9914
family3 - family6 1 3e-04 2e-04 340.5445 1.5113 0.8498
family3 - family7 1 2e-04 2e-04 340.5445 1.0294 0.9828
family3 - family8 1 4e-04 2e-04 340.5445 1.5891 0.8104
family3 - family9 1 4e-04 2e-04 340.5445 1.6034 0.8026
family5 - family6 1 1e-04 2e-04 340.5445 0.5863 0.9997
family5 - family7 1 0e+00 2e-04 340.5445 0.1045 1.0000
family5 - family8 1 2e-04 2e-04 340.5445 0.6641 0.9992
family5 - family9 1 2e-04 2e-04 340.5445 0.6784 0.9990
family6 - family7 1 -1e-04 2e-04 340.5445 -0.4818 0.9999
family6 - family8