Skip to contents

Logistic fit

Source: R/fit_curves.R
logisticFit.Rd

Fit a logistic curve to drug response data.

Usage

logisticFit(
  concs,
  norm_values,
  std_norm_values = NA,
  x_0 = 1,
  priors = NULL,
  lower = NULL,
  range_conc = c(0.005, 5),
  force_fit = FALSE,
  pcutoff = 0.05,
  cap = 0.1,
  n_point_cutoff = 4,
  capping_fold = 5
)

Arguments

concs

concentrations that have not been transformed into log space.

norm_values

normalized response values (Untreated = 1).

std_norm_values

std of values.

x_0

upper limit. Defaults to 1. For co-treatments, this value should be set to NA.

priors

numeric vector containing starting values for all. mean parameters in the model. Overrules any self starter function.

lower

numeric vector of lower limits for all parameters in a 4-param model.

range_conc

range of concentration for calculating AOC_range.

force_fit

boolean indicating whether or not to force a parameter-based fit.

pcutoff

numeric of pvalue significance threshold above or equal to which to use a constant fit.

cap

numeric value capping norm_values to stay below (x_0 + cap).

n_point_cutoff

integer indicating number of unique concentrations required to fit curve.

capping_fold

Integer value of the fold number to use for capping IC50/GR50. Default is 5.

Value

data.table with metrics and fit parameters.

Details

Implementation of the genedata approach for curve fit: https://screener.genedata.com/documentation/display/DOC21/Business-Rules-for-Dose-Response-Curve-Fitting,-Model-Selection,-and-Fit-Validity.html #nolint

The output parameter names correspond to the following definitions:

x_mean

The mean of a given dose-response metric

x_AOC_range

The range of the area over the curve

x_AOC

The area over the GR curve or, respectively, under the relative cell count curve, averaged over the range of concentration values

xc50

The concentration at which the effect reaches a value of 0.5 based on interpolation of the fitted curve

x_max

The maximum effect of the drug

ec50

The drug concentration at half-maximal effect

x_inf

The asymptotic value of the sigmoidal fit to the dose-response data as concentration goes to infinity

x_0

The asymptotic metric value corresponding to a concentration of 0 for the primary drug

h

The hill coefficient of the fitted curve, which reflects how steep the dose-response curve is

r2

The goodness of the fit

x_sd_avg

The standard deviation of GR/IC

fit_type

This will be given by one of the following:

  • "DRC4pHillFitModel" Successfully fit with a 4-parameter model

  • "DRC3pHillFitModelFixS0" Successfully fit with a 3-parameter model

  • "DRCConstantFitResult" Successfully fit with a constant fit

  • "DRCTooFewPointsToFit" Not enough points to run a fit

  • "DRCInvalidFitResult" Fit was attempted but failed

maxlog10Concentration

The highest log10 concentration

N_conc

Number of unique concentrations

Examples

logisticFit(
c(0.001, 0.00316227766016838, 0.01, 0.0316227766016838),
c(0.9999964000144, 0.999964001439942, 0.999640143942423, 0.996414342629482),
rep(0.1, 4),
priors = c(2, 0.4, 1, 0.00658113883008419)
)
#>       x_mean        x_AOC x_AOC_range  xc50     x_max      ec50     x_inf   x_0
#>        <num>        <num>       <num> <num>     <num>     <num>     <num> <num>
#> 1: 0.9993262 0.0006738196 0.003303951   Inf 0.9964143 0.0187753 0.9958965     1
#>           h        r2      p_value          rss x_sd_avg               fit_type
#>       <num>     <num>        <num>        <num>    <num>                 <char>
#> 1: 3.711073 0.9998954 0.0001046336 9.435296e-10      0.1 DRC3pHillFitModelFixS0
#>    maxlog10Concentration N_conc
#>                    <num>  <int>
#> 1:                  -1.5      4