Hi Everyone !!
With my first post this year, I wish the best for everyone !!
Coming to the confusion, I have analyzed RNASeq data of leukemia with 2 conditions (one with Loss of Heterogeneity and one without it, 14 samples to each conditions). Now I made 3 gene sets based on different pathways and functions. These gene sets had quite a few genes in common.
After I ran GSEA on these gene sets (separately) I found that genes like FOSB, ATF3, MYC they are enriched in one analysis (done with Gene set 1) while non-enriched in another analysis (done with Gene set 2) while I gave the same expression sheet, same metadata(cls file)
Q.1> when is this possible ??
Q.2> some of these genes were down regulated in DGE for LOH condition but shows enrichment in GSEAnalysis. How do we explain this ?
Thanks and Regards SJB
Hello, it is hard to follow without more context. Can you show some plots or the summary stats from the analysis? What does it mean these genes being enriched? Is your signature only three genes, or is this the leading edge genes?
Hello,
My apologies for not being able to show the plots as I am away from my computer.
Basically I have 3 gene sets (500, 350 and 700genes) and these sets have common genes like ones mentioned.
Now these genes appear "yes" in enriched column of html report page for one gene set but "no" in enriched column of html report page for second gene set. How is this possible when the expression data behind is same for both analysis.
GSEA tests whether a geneset as a whole is enriched or depleted in a comparison, not individual genes. It can be (but as said I canonly say for sure when seeing some data) that your signature is enriched which includes these genes, but some genes itself do not follow that trend. Look at this plot for example. You see that the signature as a whole is probably enriched (as most genes accumulate on the left) but some genes (on the right) escape that trend. GSEA as said checks signatures as a whole, not individual genes. Individual comparisons is what pairwise testing, e.g. DESeq2 does.