Gene set analysis - minimum number of genes for moderated t statistics

Question

[SB]

Dear Biostars Community,

Is there a generally accepted minimum number of genes recommended for a gene set in order to do moderated t-statistics for probes in the set? Would moderated t-statistics serve any advantage over t-statistics if there are only 4 genes in a set?

Thanks,

SB

Answer 1

[Gordon Smyth]

There is no minimum number of genes for a gene set. The moderation depends on all the genes in the whole data set (i.e., all the genes on the microarray or profiled by RNA-seq), not just on the genes in the test set. Moderation will give an advantage for any sized set.

PS. The limma developers answer questions on the Bioconductor Support site, so you'll get a quicker answer if you post questions about limma to that forum.