Filter Out Lowly-Correlated Genes from Signature(s)

This function removes signature genes whose expression level is poorly correlated to the signature score. This is relevant if the signatures themselves were defined from a different dataset, as might be the case in applying single-cell signatures to bulk, or in moving between datasets generated with different protocols (e.g. single-nuclei RNA seq. vs single-cell RNA seq.).

filter_signatures(m, sigs, filter.threshold = 0.4)

Arguments

m

expression matrix of genes X cells to score. Not centered.

sigs

list of gene signatures to be refined.

filter.threshold

minimum pearson's r below which genes are filtered out of the signature. Default: 0.4

Value

a filtered list of signatures, whose genes' expression levels all had correlation values to the signature score above or equal to <filter.threshold>.