Filter: Filters uninformative gene expressions during first stage of...

In DeMix: Deconvolution Models for Mixed Transcriptomes from Heterogeneous Tumor Samples

Description

To avoid unnecessary computational burdens and noises, pre-screening is an essential step. This screening process consists of three steps:

1. step 1

2. step 2

3. step 3

Usage

DeMix.Filter(input.mat, design, zerofilter=TRUE, conc=0.8, fc=1.2)

Arguments

input.mat	numeric matrix of gene expressions after normalization. The G*S matrix where G is the number of genes and S is the number of samples.
design	integer vector of S length specifying tissue components, where 0 represents healthy and 1 represents tumor samples
zerofilter	logical scalar. If TRUE, uses the screening based on zero counts.
conc	numeric scalar cut-off value used for checking the linearity assumption. More than 0.8 or 80% is recommended.
fc	numeric scalar fold-change cut-off value used for identifying informative genes in proportion estimations. 1.2 or 1/1.2 is default. This value needs to be set for reducing data-size. Approximately 2,000-3,000 genes/probes will provide robust estimates.

Value

Returns numeric G'*S} matrix of gene expressions after the prescreening, where \code{G' is the number of remaining genes and S is the number of samples.

Examples

data(ntot_NT_Liver)

# First 50 samples normal, remaining 151 tumor for this dataset
input.mat <- as.matrix(ntot_NT_Liver.df)
design <- c(rep(0, 50),
            rep(1, 151))
filtered.mat <- DeMix:::DeMix.Filter(input.mat, design)

DeMix documentation built on May 2, 2019, 5:19 p.m.