GSEA top genes vs all platform

Question

[ra2967]

Hello,

I have done an analysis in LIMMA to obtain the differential expressed genes. However, now I have 2 questions when I perform GSEA:

• Must I order my samples by fold change (first negative and after positive forld change, por example) in a preranked analysis? Is it better to order by p.val? Can I give the list obtained by limma directly in GSEA?

• GSEA is usually done with the top genes (let's say, the top 50-100 genes) or must I use THE WHOLE platform (all the 30K probes)?

Thanks in advance

[Prakash]

Have a loot at this thread, may be helpful

[glocke01]

You can also use gene set tests included in the limma package, i.e. CAMERA and ROAST. CAMERA, in particular, is way, way faster than GSEA. Prof. Smyth has made the MSigDB gene sets available in easy to ingest .rda files at his website: http://bioinf.wehi.edu.au/software/MSigDB/index.html

Answer 1

[Charles Warden]

GSEA is typically performed on the whole platform, without the differential expression step.

However, you are correct that a pre-ranked list is another option.

Answer 2

[Rashedul Islam]

Instead of using pre-ranked with 50-100 genes, you can use entire set. Then check how many genes have core GSEA enrichment and are differentially expressed.