processSeq: Process Sequencing Data for Poisson-based MRFs

In XMRF: Markov Random Fields for High-Throughput Genetics Data

Description

Process and normalize RNA-Sequencing count data into a distribution appropriate for Poisson MRFs.

Usage

processSeq(X, quanNorm = 0.75, nLowCount = 20, percentLowCount = 0.95, NumGenes = 500, 
PercentGenes = 0.1)

Arguments

X	nxp data matrix.
quanNorm	an optional parameter controlling the quantile for sample normalization, default to 0.75.
nLowCount	minimum read count to decide if to filter a gene, default to 20.
percentLowCount	filter out a gene if it has this percentage of samples less than nLowCount, default to 0.95.
NumGenes	number of genes to retain in the final data set, default to 500.
PercentGenes	percentage of genes to retain, default to 0.1.

Details

To process the next-generation sequencing count data into proper distribution (with dispersion removed), the following steps are taken in this function:

1. Quantile normalization for the samples.

2. Filter out genes with all low counts.

3. Filter genes by maximal variance (if specified).

4. Transform the data to be closer to the Poisson distribution. A log or power transform is considered and selected based upon the Kolmogorov-Smirnov goodness of fit test.

Value

a n x NumGenes or PercentGenes processed data matrix.

Examples

library(XMRF)
data('brcadat')
brca = t(processSeq(t(brcadat), PercentGenes=1))

Example output

XMRF documentation built on May 2, 2019, 8:18 a.m.