Entering edit mode
23 months ago
jabbari.parnian
▴
30
I'm using GSEA for my RNA-Seq data to find significant pathways. The software instructs to use normalized count data so I used TMM normalization by edgeR. However, I don't get any significant gene sets based on FDR. What would be the problem if I use non-normalized counts with just low counts filtered out? Thank you.
What exactly is your experiment and goal?
finding gene sets that are changed between the two experimental groups.
Not sure if the normalization method has an impact on GSEA results, but you can give it a try also by normalizing with DESeq2. Usually, that's what I do.
However, how many DEGs have you identified with your pipeline?
Using DESeq2, I found 138 upregulated genes and 118 downregulated genes. I'll try normalization with DESeq2 as well. Thanks.
how does the PCA of the normalized counts look? can you see the grouping of samples as per metadata/variables?
I haven't performed that yet, but I'll try. Thanks
QC like that should be done first, be sure to do that. Also, one would need details which GSEA implementation you did run and how you normalized data.