tn.norm: Targeted-normalization of bin counts

In jmonlong/PopSV: Population-based detection of structural variants from Read-Depth signal

Description

Bin counts are normalized one bin at a time, using a subset of the bins that look similar across the reference samples.

Usage

tn.norm(bc, cont.sample, nb.support.bins = 1000, bins = NULL,
  save.support.bins = TRUE, norm = c("1pass", "trim"),
  force.diff.chr = TRUE)

Arguments

bc	a matrix or data.frame with the bin counts (bin x sample).
cont.sample	the sample to use as baseline for the pairwise normalization. All the samples will be normalized to it.
nb.support.bins	the number of bins to use for the normalization.
bins	a vector the names of the bins to normalize. If NULL (default), all bins are normalized.
save.support.bins	if TRUE (default) the bins used for the normalization are saved in the output object 'norm.stats'.
norm	the type of normalization. '1pass' (default) means one pass of normalization. Other options is 'trim' (/!\ experimental /!\).
force.diff.chr	should the supporting bins be forced to be in a different chromosome. Default is TRUE.

Details

A specific set of bins, defined by 'bins=', can de normalized using all the bins in 'bc'. The bin names in 'bins=' should be the chr and start position as in '1-501' (for chr 1, start 501).

The default approach ('norm="1pass"') looks for supporting across all samples. A more robust approach, finds supporting bins after trimming a few outlier samples (potential CNV). 'norm="trim"' normalization is better and recommended for small bins.

Value

a list with

norm.stats	a data.frame witht some metrics about the normalization of each bin (row) : correlation with worst supporting bin ('d.max'); coverage average ('m') and standard deviation ('sd'); number of outlier reference samples ('nb.remove'); supporting bins.
bc.norm	a matrix with the normalized bin counts (bin x sample).
nb.support.bins, cont.sample	a backup of the input parameters.

Author(s)

Jean Monlong

jmonlong/PopSV documentation built on Sept. 15, 2019, 9:29 p.m.