Hi,

I've been using the shrunken log fold-changes generated by apeglm as the test statistic for gene set enrichment analyses when running the fgsea package. I have seen a paper that suggests the absolute value of Signal-To-Noise is a better metric than LFC: https://bmcbioinformatics.biomedcentral.com/articles/10.1186/s12859-017-1674-0 However, the study was performed using microarray datasets; hence as you can calculate S/N from the output of apeglm (shrunken LFC / standard error of shrunken LFC), I was wondering if anyone has tried this approach with RNAseq datasets please? Would this be "valid" as the LFCs have been shrunk & may not be appropriate for calculating S/N ratios?

Many thanks for any comments/ideas.

Cheers, Richard.

I would recommend to take just the shrunken LFC. If there is a lot of noise in the data, these estimates are moved toward zero. It qualitatively is similar to S/N except I think has the advantage that a gene with a small S but an even smaller N will not be blown up by lfcShrink.