smoothBSCounts: Dirichlet smoothing of BS-seq counts prior to compartment...

In JordanVeldboom/compartmap: A/B compartment inference from ATAC-seq and methylation array data

Description

helper function for compartment inference (shrink-by-smoothing logit frac5mC)

We want something with nominally Gaussian error for compartment inference, so this function grabs suitable (default >= 3 reads in >=2 sample) measurements and turns them into lightly moderated, logit-transformed methylated-fraction estimates (also known, unfortunately, as M-values) for compartment calling, by performing Dirichlet smoothing (adding k reads to M and U support).

Usage

smoothBSCounts(obj, minCov = 3, minSamp = 2, k = 0.5, r = NULL)

Arguments

obj	a BSseq object with methylated and total reads
minCov	minimum read coverage for landmarking samples (3)
minSamp	minimum landmark samples with >= minCov (2)
k	pseudoreads for smoothing (0.5)
r	regions to collapse over (default is NULL, do it by CpG)

Value

       smoothed logit(M/Cov) matrix with coordinates as row names

JordanVeldboom/compartmap documentation built on July 3, 2020, 6:32 p.m.