Many DESeq2 P values are 0 thus preventing generation of a rank list for GSEA

Question

[Krish Srinivasan]

Hi Everyone,

In my DESeq2 reuslts files, the smallest 100 P-values are 0 (as the smallest floating point value in R is 2.225074e-308). I am wondering if there is any way to discern the unrounded values of these P-values as they are needed in order for me to produce a rank list for gene set enrichment analysis (GSEA). Any guidance on how to approached ranking these genes would be appreciated.

Thank you

[h.mon]

Quick and dirty: use log-fold(change) instead. Or rank by P value, then by log-fold(change).

P.S.: see discussion here for alternatives, implications and explanations.

Answer 1

[Martombo]

you can use the stat column in the results object, as this (in absolute values) is monotonic when compared to the p-values. Some additional info on what metric to use for ranking can be found in this paper.

[Krish Srinivasan]

Thanks for the article, I'm sure it will be very helpful!