Skip to contents

fit_SE for time-course (Incucyte) screens

Source: R/fit_SE.timecourse.R
fit_SE.timecourse.Rd

Two-stage pipeline for time-course dose-response data:

  1. Compute growth rates per user-defined time window and normalise them to the control of the reference period (compute_growth_rates).

  2. Fit a 3-parameter Hill (logistic) curve on the NormalizedGrowthRate ~ log10(Concentration) dose-response for each (Drug, CellLine, period) combination.

Usage

fit_SE.timecourse(
  se,
  periods,
  normalization_map,
  lfc_assay = NULL,
  metrics_assay = "Metrics",
  rate_fn = NULL,
  fit_fn = NULL,
  fit_source = NULL,
  n_point_cutoff = 4L,
  range_conc = c(0.005, 5),
  force_fit = FALSE,
  pcutoff = 0.05,
  cap = 0.1
)

Arguments

se

SummarizedExperiment with a LogFoldChange assay (output of normalize_SE for data_type = "time-course").

periods

named list of numeric vectors of length 2; each element defines a half-open time window c(start_hour, end_hour) used for growth-rate calculation: a timepoint is used when start_hour <= Duration < end_hour, so contiguous windows partition the time-course without dropping the shared boundary. At minimum periods$early must be provided. Example: list(early = c(0, 48), mid = c(48, 96), late = c(96, 144)).

normalization_map

named character vector mapping every period name to the reference period for DMSO normalization, or "None" to keep the raw growth rate. Its names must match names(periods) exactly and its values must be "None" or an existing period name. Example: c(early = "None", mid = "early", late = "early").

lfc_assay

string; name of the log-fold-change assay in se. Defaults to the input_assay of the "time-course" fit profile.

metrics_assay

string; name of the output metrics assay. Default "Metrics".

rate_fn

function or NULL; growth-rate estimator passed to compute_growth_rates. NULL uses the default lm(LogFoldChange ~ Duration) slope.

fit_fn

function or NULL; dose-response fit applied to each normalization_type slice of the GrowthRates assay. NULL uses fit_drug_response_metrics on NormalizedGrowthRate. When a custom function is supplied, n_point_cutoff, range_conc, force_fit, pcutoff and cap are not used.

fit_source

string or NULL; value recorded in the fit_source column. NULL records "gDR" for the default fit and "custom" for a user-supplied fit_fn.

n_point_cutoff

integer; minimum number of unique concentrations required to attempt a curve fit. Default 4L.

range_conc

numeric vector of length 2; concentration range for x_AOC_range. Default c(5e-3, 5).

force_fit

logical; if TRUE fit is kept even if not statistically significant. Default FALSE.

pcutoff

numeric; p-value threshold for the F-test. Default 0.05.

cap

numeric; upper-capping fold above x_0 = 1. Default 0.1.

Value

A new SummarizedExperiment where rows = (Drug, CellLine) and columns = period names, holding the GrowthRates assay from stage 1 and the Metrics assay (or metrics_assay) from stage 2. With the default fit_fn the metrics are: ec50, xc50, h, x_inf, x_0, r2, x_mean, x_AOC, x_AOC_range, x_max, fit_type, N_conc, maxlog10Concentration, and period.

Details

The LogFoldChange assay consumed by this function is produced by normalize_SE when data_type = "time-course".

Both stages accept a replacement function, so custom fitting works for time-course data the same way it does for single-agent and combination screens. The two stages can also be run separately: compute_growth_rates returns the growth rate table (controls included) and growth_rates_to_se turns it into the input of apply_fit with data_type = "time-course-metrics".

See also

compute_growth_rates for stage 1 on its own, fit_SE for single-agent fitting, fit_SE.combinations for combination screens, apply_fit for the generic fitting engine.

Examples

if (FALSE) { # \dontrun{
# After normalize_SE(se_tc, data_type = "time-course"):
periods <- list(early = c(0, 48), mid = c(48, 96), late = c(96, 144))
norm_map <- c(early = "None", mid = "early", late = "early")
se_fit <- fit_SE.timecourse(se_tc, periods = periods,
                            normalization_map = norm_map)
gDRutils::convert_se_assay_to_dt(se_fit, "Metrics")

# Custom growth-rate estimator and custom dose-response fit:
se_custom <- fit_SE.timecourse(
  se_tc,
  periods = periods,
  normalization_map = norm_map,
  rate_fn = function(dt) 24 * stats::median(diff(dt$LogFoldChange) /
                                            diff(dt$Duration)),
  fit_fn = function(dt) data.table::data.table(x_mean = mean(dt$NormalizedGrowthRate))
)
} # }