I have a query regarding the analysis of GSEA Results. For GSEA, I have used "fgsea" R package to obtain the dysregulated KEGG pathways. Now, I want to rank the dysregulated KEGG pathways.
I have seen people ranking the dysregulated genes on the basis of log2Fold Change and p-value obtained from DESeq2. For genes, they use the decreasing order of log2FC * (-log10 p-value) to rank the genes.
So, is it logical to use NES * (- log10 Nominal p-value) for ranking the KEGG pathways?
Edit: The idea was inspired by people ranking the DEGs for preranked GSEA using their log2FC * (-log10 p-value). Also, the GSEA results I am getting, don't have any pathway with adjusted p-value <0.25. So, I can't rank the results by absolute NES after taking cutoff by p-adjusted value.
Where did you see ranking of genes by
log2FC * (-log10 p-value)? I don't think I ever saw one and not sure what it should achieve. Aren't you confusing it with ranking bysign(log2FC) * (-log10 p-value)?You can rank the output of GSEA by the NES there is no need to use the p-value for ranking. Why do you think it is needed?
The idea was inspired by people ranking the DEGs for preranked GSEA using their log2FC * (-log10 p-value). Also, the GSEA results I am getting, don't have any pathway with adjusted p-value <0.25. So, I can't rank the results by absolute NES after taking cutoff by p-adjusted value.
You can have the ranking without the cutoff. If you don't keep any gene set after the cutoff then there is nothing to rank as you said, so using log10(p-value) won't solve your problem.