Hi all,

I have done some drug response profiling. There are some drugs that are only effective in one or two samples (around 20 screened). I also have some RNA-Seq data from these samples (plus 50 more without screening results) and I calculated the GSVA score for them with rlog output from DESeq2.

I am just wondering if it is a legit way of finding activated pathway for a SINGLE sample by calculating the Z-score of GSVA normalised across samples per gene set and pick those with Z-score higher than 1.96.

I appreciate it if there is a more legit way of doing it and you can let me know. Thanks!

Hi, if you want to do some formal inference about the effect of a drug at RNA level, then I would recommend you doing a differential expression analysis on the GSVA scores derived from RNA-seq data, between the group of samples that were exposed to the drug and the group of samples that were not exposed to the drug. I would recommend running gsvaParam() and gsva() with default parameters. Please consult the subsection on "Differential expression at pathway level" from the GSVA vignette.