Skip to contents

Dose-Response Analysis

Source: vignettes/dose-response.Rmd
dose-response.Rmd

R-CMD-check pkgdown CRAN status Codecov test coverage CRAN downloads CRAN downloads total License: MIT Lifecycle: experimental

cellreportR fits three standard dose-response models:

  • 4PL (four-parameter log-logistic)
  • 3PL (fixed lower asymptote at zero)
  • Linear (for convenience; fits y ~ x)

Setting up a dose-response experiment

Any cr_experiment with a dose column in design is eligible.

exp <- cr_example_experiment(seed = 3, n_cells_per_well = 60)
exp$design$dose <- dplyr::case_when(
  exp$design$treatment == "Untreated"      ~ 0.1,
  exp$design$treatment == "CompoundA_low"  ~ 50,
  exp$design$treatment == "CompoundA_high" ~ 500,
  exp$design$treatment == "PosControl"     ~ 1000,
  TRUE ~ 10
)

Fitting 

fit <- cr_dose_response(exp,
                        channel = "marker_1",
                        model = "4pl",
                        log_dose = TRUE)
fit$model
#> [1] "4pl"
fit$params
#> # A tibble: 4 × 3
#>   parameter estimate std_error
#>   <chr>        <dbl>     <dbl>
#> 1 a           615.          NA
#> 2 d          4503.          NA
#> 3 e             8.90        NA
#> 4 b         -3072.          NA

IC50 / EC50 

cr_ic50(fit)
#> # A tibble: 1 × 5
#>   parameter   estimate ci_low ci_high units
#>   <chr>          <dbl>  <dbl>   <dbl> <chr>
#> 1 IC50      788211529.     NA      NA uM

Plotting

Troubleshooting

If nls() fails to converge (for example when there are very few dose levels), cellreportR falls back to a linear model. Check fit$model after fitting.