rebin_matrix: Transforms a bins x cells count matrix into a larger bins x...

In vallotlab/ChromSCape: Analysis of single-cell epigenomics datasets with a Shiny App

Transforms a bins x cells count matrix into a larger bins x cells count matrix.

Description

This functions is best used to re-count large number of small bins or peaks (e.g. <= 5000bp) into equal or larger sized bins. The genome is either cut in fixed bins (e.g. 50,000bp) or into an user defined number of bins. Bins are calculated based on the canconical chromosomes. Note that if peaks are larger than bins, or if peaks are overlapping multiple bins, the signal is added to each bin. Users can increase the minimum overlap to consider peaks overlapping bins (by default 150bp, size of a nucleosome) to disminish the number of peaks overlapping multiple region. Any peak smaller than the minimum overlapp threshold will be dismissed. Therefore, library size might be slightly different from peaks to bins if signal was duplicated into multiple bins or ommitted due to peaks smaller than minimum overlap.

Usage

rebin_matrix(
  mat,
  bin_width = 50000,
  custom_annotation = NULL,
  minoverlap = 500,
  verbose = TRUE,
  ref = "hg38",
  nthreads = 1,
  rebin_function = rebin_helper
)

Arguments

mat	A matrix of peaks x cells
bin_width	Width of bins to produce in base pairs (minimum 500) (50000)
custom_annotation	A GenomicRanges object specifying the new features to count the matrix on instead of recounting on genomic bins. If not NULL, takes predecency over bin_width.
minoverlap	Minimum overlap between the original bins and the new features to consider the peak as overlapping the bin . We recommand to put this number at exactly half of the original bin size (e.g. 500bp for original bin size of 1000bp) so that no original bins are counted twice. (500)
verbose	Verbose
ref	Reference genome to use (hg38)
nthreads	Number of threads to use for paralell processing

Value

A sparse matrix of larger bins or peaks.

Examples

mat = create_scDataset_raw()$mat
binned_mat = rebin_matrix(mat,bin_width = 10e6)
dim(binned_mat)

vallotlab/ChromSCape documentation built on Oct. 15, 2023, 1:47 p.m.