How to rank genes prepared for GSEA analysis with TopGO

Question

[wangdp123]

Hi there,

I am trying to use TopGO to perform some GSEA analyses (Gene Set Enrichment Analysis). In order to do this, I have done the differential expression analysis for all genes and produced the P-values and Fold change values for each gene.

I am aware of the fact that TopGO needs a ranked gene list based on only one metric. As I understand, the top of the list should be the most up-regulated genes and the bottom being the most down-regulated ones and moderate genes will be located in the middle part. The significance of the difference is reflected on two metric: P-values and Fold changes. My intuition tells me that only one is not enough.

Any thought about this? how to combine these two metrics into one or any other good idea?

Many thanks,

Tom

Answer 1

[BioinfGuru]

You are correct in "the significance of the difference is reflected by two metrics" but they are p-value and FDR (or p-value and q-value depending on the program you use).

Fold change is actually the difference between the expression values between the 2 groups. If you used deseq then fold = concSample1 - concSample2.

The metric you need to use to rank the gene list is the FDR or q-value whichever of those is given. Don't use the fold change as even a large fold change may not be significant (for many reasons not needed here). Don't use p-value because it is not corrected for multiple testing.