Hi Robert,

I am new to GSVA package. I am trying to the which geneset is differentially expressed in my samples? So, far i tried taking my differentially expressed genes as my input and my geneSets as reference file and ran gsva function opting "gsva" method of gsva() function. Next, to find which differentially expressed geneSets i am using limma approach and considering significant geneSets if corrected p value (FDR) is less than 0.2. Am i performing it correctly?

When I am using gsva function, there are three different methods we can use like "gsva", "ssgsea" and "zscore". Which one is more stringent way to use or which method is widely used one. I tried ssgsea and gsva both is giving me different results. I am just wondering which one to use?

If i use gsva method, it results me in both positive and negative enrichment scores if i use ssgsea method sometimes it yields me only positive values where gsva gave me both positive and negative enrichment scores. I am little apprehensive whether or not i am using the function correctly.

Any help would be really appreciated.

Our consortium have been playing around with GSVA and I found that a tool called QuSAGE is more useful. This is because you can use the Qgen function to correct for covariates in a linear model whereas GSVA does not allow this. It also corrects for inter gene correlations which GSVA does not do. This isn't strictly answering your question, but in my view GSVA is inferior to QuSAGE, I also found it straightforward to use, I remember having a few issues with GSVA.

How does the QuSAGE take into account the covariates? I am used to GSVA and after calculating the scores one can use them with the normal workflow of limma, thus taking into account covariates to estimate the fold-change of pathways. Although the plots of QuSAGE seem appealing