chordPredict: Predict the probability of homogolous recombination...

In luannnguyen/CHORD: Classifier of Homologous Recombination Deficiency

Predict the probability of homogolous recombination deficiency using mutational signatures

Description

A wrapper for predict.randomForest() from the randomForest package

Usage

chordPredict(
  features,
  rf.model = CHORD,
  hrd.cutoff = 0.5,
  trans.func = NULL,
  min.indel.load = 50,
  min.sv.load = 30,
  min.msi.indel.rep = 14000,
  do.bootstrap = F,
  bootstrap.iters = 20,
  bootstrap.quantiles = c(0.05, 0.5, 0.95),
  detailed.remarks = T,
  show.features = F,
  verbose = T
)

Arguments

features	The output of extractSigsChord(), which is a dataframe containing the SNV, indel and SV context counts.
rf.model	The random forest model. Defaults to CHORD.
hrd.cutoff	Default=0.5. Samples greater or equal to this cutoff will be marked as HRD (is_hrd==TRUE).
trans.func	Function used to transform raw features. Raw features should be in the format: list(snv=matrix(), indel=matrix(), sv=matrix())
min.indel.load	Default=50. The minimum number of indels required to make an accurate HRD prediction. Samples with fewer indels than this value will be marked as is_hrd==NA (HR status could not be confidently determined).
min.sv.load	Default=30. The minimum number of SVs required to make an accurate prediction of BRCA1-type vs. BRCA2-type HRD. Samples with fewer SVs than this value will be marked as hrd_type==NA (HRD type could not be confidently determined).
min.msi.indel.rep	Default=14000 (changing this value is not advised). Samples with more indels within repeats than this threshold will be considered to have microsatellite instability.
do.bootstrap	Test the stability of prediction probabilities? NOTE: this is computationally expensive. Resamples the feature vector for each sample (number of times provided by bootstrap.iters) and calculates HRD probabilities for each iteration. Returns the probabilities at the quantiles specifying in bootstrap.quantiles
bootstrap.iters	Number of resampling iterations for determining the confidence intervals
bootstrap.quantiles	A numeric vector of length 2 specifying the quantiles used to calculate the confidence intervals
detailed.remarks	If TRUE, shows min.indel.load and min.sv.load numbers in the remarks columns
show.features	If TRUE, appends features to output
verbose	Show messages?

Value

A dataframe containing the HRD probabilities, bootstrap probabilities, and input features

Examples

## Extract mutation contexts
vcf_dir <- '/path_to_vcfs/'
vcf_snv <- paste0(vcf_dir,'SampleX_post_processed_v2.2.vcf.gz')
vcf_indel <- paste0(vcf_dir,'SampleX_post_processed_v2.2.vcf.gz')
vcf_sv <- paste0(vcf_dir,'SampleX_somaticSV_bpi.vcf.gz')
contexts <- extractSigsChord(vcf_snv, vcf_indel, vcf_sv, sample.name='SampleX')

## Predict HRD probability with CHORD
chordPredict(contexts)

luannnguyen/CHORD documentation built on Aug. 25, 2023, 10:04 a.m.