Skip to contents

Voom-transform a count matrix

rs_voom_normalise.Rd

[Experimental] limma's voom on counts that are already filtered: log2-CPM against the supplied (effective) library sizes, the mean-variance trend and the precision weights. No filtering and no normalisation happen in here; pass lib.size * norm.factors as lib_size to get voom on a normalised DGEList.

Usage

rs_voom_normalise(counts, design, lib_size, span, adaptive_span)

Arguments

counts

Integer or double matrix. Raw counts of genes x samples.

design

Numeric matrix. The design matrix of samples x coefficients. Must be full rank.

lib_size

Numeric vector. The effective library size per sample.

span

Numeric. Lowess span, only used if adaptive_span = FALSE.

adaptive_span

Boolean. Derive the span from the number of genes, as limma does since 3.56.

Value

A list with the following elements

  • e - Numeric matrix. The log2-CPM values, genes x samples. limma's E.

  • weights - Numeric matrix. The precision weights, genes x samples.

  • trend_x - The mean-variance trend abscissae (average log2 count).

  • trend_y - The mean-variance trend ordinates (sqrt standard deviation).

  • amean - Average log2-CPM per gene.

References

Law, et al., Genome Biol, 2014