ELISA curve fitting using ggplot2

Introduction

Here is an example of how to fit and analyse ELISA data using ggplot2, drc, and broom.

library(tidyplate)
library(tidyverse)
library(drc)
library(broom)
library(ggsci)

First we import plate data using the tidy_plate function from the tidyplate package:

raw_data <- tidy_plate("data/elisa_example.xlsx")
glimpse(raw_data)

Rows: 50
Columns: 10
$ well                <chr> "A01", "A02", "A03", "A04", "A05", "A06", "B01", "…
$ coat_protein_name   <chr> "sBACE", "sBACE", "sBACE", "sBACE", "sBACE", "sBAC…
$ coat_protein_ug     <dbl> 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, 0.2, …
$ coat_protein_source <chr> "NS0", "NS0", "NS0", "NS0", "NS0", "NS0", "NS0", "…
$ primary_mab_name    <chr> "aBACE", "aBACE", "aBACE", "aSARS", "aSARS", "aSAR…
$ primary_mab_clone   <chr> "6626.1", "6626.1", "6626.1", "CC6.29", "CC6.29", …
$ primary_mab_conc    <dbl> 10.0000000, 10.0000000, 10.0000000, 10.0000000, 10…
$ secondary_mab_name  <chr> "goat-aHuman", "goat-aHuman", "goat-aHuman", "goat…
$ secondary_mab_dil   <chr> "1to5000", "1to5000", "1to5000", "1to5000", "1to50…
$ od450               <dbl> 1.396, 1.170, 1.299, 1.324, 1.170, 1.299, 1.374, 1…

Prepare data before plotting

blank_data <- raw_data |>
    dplyr::filter(primary_mab_name == "blank")
mean_blank <- mean(blank_data[["od450"]], na.rm = TRUE)

summary <- raw_data |>
    dplyr::filter(primary_mab_name != "blank") |>
    dplyr::mutate(blanked_od = od450 - mean_blank) |>
    dplyr::group_by(primary_mab_name, primary_mab_conc) |>
    dplyr::summarise(
      mean_od = mean(blanked_od, na.rm = TRUE),
      mean_sd = sd(blanked_od, na.rm = TRUE),
      .groups = 'drop'
    )
head(summary)

# A tibble: 6 × 4
  primary_mab_name primary_mab_conc mean_od mean_sd
  <chr>                       <dbl>   <dbl>   <dbl>
1 aBACE                     0.00244   0.138  0.0797
2 aBACE                     0.00977   0.214  0.0626
3 aBACE                     0.0391    0.395  0.0644
4 aBACE                     0.156     0.740  0.0531
5 aBACE                     0.625     1.11   0.0358
6 aBACE                     2.5       1.28   0.0486

The final plot

theme_set(theme_bw(base_size = 20))
ggplot(summary, aes(x = primary_mab_conc, 
                    y = mean_od, 
                    group = primary_mab_name, 
                    color = primary_mab_name)) +
  geom_point(shape = 21, size = 3, stroke = 1.5) +
  geom_smooth(method = drc::drm, 
              method.args = list(fct = drc::L.4()),
              se = FALSE, linewidth = 1) +
  geom_errorbar(aes(ymin = mean_od - mean_sd,
                    ymax = mean_od + mean_sd),
                width = 0.1) +
  scale_x_log10() +
  scale_color_npg() +
  labs(x = "Log of antibody concentration (µg/ml)",
       y = "OD~450~",
       color = NULL) +
  theme(axis.title.y = ggtext::element_markdown(),
        legend.position = c(0.2, 0.7))

Statistical analysis of the fit

We have to remove “blanks” first before modeling.

raw_data_no_blanks <- subset(raw_data, primary_mab_name != "blank")
drm_model <- drm(formula = od450~primary_mab_conc, 
                 curveid = primary_mab_name, 
                 data = raw_data_no_blanks, 
                 fct = LL.4(names=c("Slope", "Lower", "Upper", "ED50")))
# tidy(drm_model)
# glance(drm_model)
# augment(drm_model, data = raw_data_no_blanks)

ggplot(augment(drm_model, data = raw_data_no_blanks), 
       aes(.fitted, .resid, color = as.factor(primary_mab_conc))) +
  geom_hline(yintercept = 0) +
  geom_point(shape = 21, size = 3, stroke = 1.5) +
  labs(color = "dilutions") +
  scale_color_npg() +
  facet_grid(cols = vars(primary_mab_name))