NBumi_CoExpression: Variance-based Feature Selection

In tallulandrews/M3Drop: Michaelis-Menten Modelling of Dropouts in single-cell RNASeq

Variance-based Feature Selection

Description

Ranks genes by residual dispersion from mean-dispersion power-law relationship.

Usage

	NBumiCoexpression(counts, fit, gene_list=NULL, method=c("both", "on", "off"))

Arguments

counts	raw count matrix (e.g. from NBumiConvertData).
fit	output from NBumiFitModel or NBumiFitBasicModel.
gene_list	set of gene names to test coexpression of.
method	type of coexpression to test (see: Details).

Details

Tests for co-expression using the normal approximation of a binomial test. Co-expression is defined according to the method argument as follows:

on

two genes are both >0 in more cells than expected.

off

two genes are both 0 in more cells than expected.

both

two genes are either both >0 or both 0 in more cells than expected.

In all cases the null expectation is calculated using the fit library-size adjusted negative binomial model. This remove spurious co-expression due to cells with differing numbers of detected genes.

Value

a matrix of Z-scores for each pair of genes in the provided gene list.

Examples

	library(M3DExampleData)
	counts <- NBumiConvertData(Mmus_example_list$data)
	fit <- NBumiFitModel(counts);
	genes <- c("Sox2", "Eomes", "Zscan4d", "Obox1", "Obox3")
	co <- NBumiCoexpression(counts, fit, genes, method="both");
	on <- NBumiCoexpression(counts, fit, genes, method="on");
	off <- NBumiCoexpression(counts, fit, genes, method="off");

tallulandrews/M3Drop documentation built on March 6, 2024, 1:49 a.m.