BRD: Background Reads Density

In benja0x40/Tightrope: ChIP-seq normalization according to background reads density

Description

The BRD function searches for background candidates based on the provided read count matrix and uses these background candidates to estimate normalization factors.

Usage

BRD(cnt, controls, sampling = NULL, smobs = NULL, dither = NULL,
  zscore = NULL, bins = NULL, smoothing = NULL, bdt = NULL,
  ncl = NULL, mincs = NULL)

Arguments

cnt	matrix of read counts (rows = observations, columns = measurement conditions).
controls	count columns corresponding to control measurements (Input, IgG, etc.).
smobs	subtract a mean count value for each observation (default = TRUE, recommended).
dither	number of replicates of the count dithering performed by DitherCounts: 1 = single, 2 = duplicate, 3 = triplicate, etc. (default = 5).
zscore	transform read count projections into z-scores (default = TRUE, recommended).
bins	number of bins per principal component for density estimations (default = 500).
smoothing	number of consecutive bins for local average smoothing of estimated densities (default = 10).
bdt	numeric vector of length 2 defining background density thresholds both expressed as proportions between 0 and 1. bdt[1] specifies the global threshold used to discard observations with low density prior to clustering. bdt[2] determines the maximum density loss allowed when selecting core observations relatively to the local maximum density in each cluster, and thus defining the background candidates. By default the value of bdt is c(0.2, 0.05) meaning that, in terms of density percentiles, the bottom 20 percents will be filtered out before clustering and only the top 5 percents can be selected as background candidates among each cluster.
ncl	number of clusters (density modes) to be distinguished.
mincs	minimum size of cluster cores, as number of observations.

Value

BRD returns a list with the following elements:

parameters	call parameters of the function.
status	execution status.
nonzero	indices of initial observations with count > 0.
dred	dimensionality-reduced non-zero observations.
subsets	partition of non-zero observations into background candidate subsets.
populations	summary of the core population in each subset.
theta	fitted distribution parameters for each core population.
log2counts	dithered and log2 transformed counts.
normfactors	BRD normalization factors.

See Also

PlotBRD, BackgroundCandidates, ScalingFactors, NormalizeCountMatrix, ReadCountMatrix

benja0x40/Tightrope documentation built on May 24, 2019, 1:35 a.m.