callingWaterfall: Drug sensitivity calling using waterfall plots
In bhklab/PharmacoGx: Analysis of Large-Scale Pharmacogenomic Data

callingWaterfall

R Documentation

Drug sensitivity calling using waterfall plots

Description

Sensitivity calls were made using one of IC50, ActArea or Amax

Usage

callingWaterfall(
  x,
  type = c("IC50", "AUC", "AMAX"),
  intermediate.fold = c(4, 1.2, 1.2),
  cor.min.linear = 0.95,
  name = "Drug",
  plot = FALSE
)

Arguments

`x`	What type of object does this take in?
`type`	ic50: IC50 values in micro molar (positive values) actarea: Activity Area, that is area under the drug activity curve (positive values) amax: Activity at max concentration (positive values)
`intermediate.fold`	vector of fold changes used to define the intermediate sensitivities for ic50, actarea and amax respectively
`cor.min.linear`	`numeric` The minimum linear correlation to require?
`name`	`character` The name of the output to use in plot
`plot`	`boolean` Whether to plot the results

Details

Sort log IC50s (or ActArea or Amax) of the samples to generate a “waterfall distribution”
Identify cutoff:

3.1 If the waterfall distribution is non-linear (pearson cc to the linear fit <=0.95), estimate the major inflection point of the log IC50 curve as the point on the curve with the maximal distance to a line drawn between the start and end points of the distribution.

3.2 If the waterfall distribution appears linear (pearson cc to the linear fit > 0.95), then use the median IC50 instead.

Samples within a 4-fold IC50 (or within a 1.2-fold ActArea or 20% Amax difference) difference centered around this inflection point are classified as being “intermediate”, samples with lower IC50s (or ActArea/Amax values) than this range are defined as sensitive, and those with IC50s (or ActArea/Amax) higher than this range are called “insensitive”.
Require at least x sensitive and x insensitive samples after applying these criteria (x=5 in our case).