unexplainedSamples: Estimate samples not fully explained by signature fit

In pdiakumis/hrdetect: Signature Tools

Description

Given a catalogue of samples, signatures and exposures, compute the sum of the absolute deviations (SAD) between the original catalogue and the reconstructed samples (i.e. signatures x exposures) and normalise this sum by the total number of mutations in the sample. Then, for each sample, compare its normalised SAD to the normalised SAD of the other samples and check if it is significantly different. In practice, a p-value is computed fitting a gaussian distribution to the other samples.

Usage

unexplainedSamples(
  fileout = NULL,
  catalogue,
  sigs,
  exposures,
  pvalue_threshold = 0.01
)

Arguments

fileout	if specified, generate a plot, otherwise no plot is generated
catalogue	original catalogue, channels as rows and samples as columns
sigs	mutational signautures used for fitting, channels as rows, signatures as columns
exposures	exposures/activities of signatures in each sample. Signatures as rows, samples as columns
pvalue_threshold	threshold for statistical significance

Value

table of unexplained samples

Examples

res <- SignatureFit_withBootstrap(cat = catalogue,
                  signature_data_matrix = cosmic30,
                  nboot = 5,
                  threshold_percent = 0.1,
                  threshold_p.value = 0.1)
s_table <- unexplainedSamples(catalogue=catalogue,
                  sigs=cosmic30,
                  exposures=res$E_median_filtered)

pdiakumis/hrdetect documentation built on May 17, 2020, 5:30 p.m.