Hello,

I have inherited a table of differential expression results from edgeR (with logFC, logCPM, LR, PValue and FDR). Unfortunately, I do not have access to the original gene counts.

I would like to perform some gene set testing using this data with the Broad MSigDB gene set collections. I was wondering whether the pre-ranked version of camera could work, either with the logFC alone or the LR multiplied by the sign of the logFC as the statistic argument for cameraPR.

I do not really like the idea of the first option but the second does not seem ideal either. If they are entirely non-sensical, are there any better alternatives?

Thanks.

Hi Jamie,

Yes, you can use cameraPR() with the edgeR results. As test statistic, the signed likelihood ratio test statistic would be best:

statistic <- sign(logFC) * sqrt(LR)