which set of results are more reliable and precise for DEG analysis, limma vs GSEA ?

Question

[Shamim Sarhadi]

I have done DEG analysis with limma and GSEA separately ,in 150 topranked of each one, I have different results, when I run Reduce(intersect, list(limma ,GSEA)) I found that, there are 43 common gene symbols

Now I don't know which results is more reliable for downstream analyze and please let me know,in your opinion, what is the source of this discrepancy ? I should tell you that before I keep prob-ids with Amean > median, these common gene were 25 in 150 topranked list !!

Answer 1

[ivivek_ngs]

I would like to highlight that GSEA is not a tool for differential expression analysis. You can check what is written in the website

*

In particular: Prior to conducting gene set enrichment analysis, conduct your differential expression analysis using any of the tools developed by the bioinformatics community (e.g., cuffdiff, edgeR, DESeq, etc). Based on your differential expression analysis, rank your features and capture your ranking in an RNK-formatted file. The ranking metric can be whatever measure of differential expression you choose from the output of your selected DE tool. For example, cuffdiff provides the (base 2) log of the fold change. Run GSEAPreranked, but make sure to select "classic" for your enrichment score (thus, not weighting each gene's contribution to the enrichment score by the value of its ranking metric).

*

It is reccomended to use the different DEG tools to find out the genes that are differentially expressed and then use the GSEAPreRanked. I believe you should try count based tools like DESeq2 , edgeR with the same normalization method and try to find the overlap of them and then perform your downstream or rather use any of them to go ahead with GSEA.

Please take a look at this link section

Can I use GSEA to analyze SNP, SAGE, ChIP-Seq or RNA-Seq data?

I believe this will be able to make it more clear for you.

Answer 2

[andrew.j.skelton73]

This obviously depends on your experimental design, and how you're using the tools, but those are two completely different methods. If you're looking for strict differential expression, i.e. comparing the means of two phenotypic states (in a nutshell), then limma is the tool that you want to use. GSEA uses a priori information (genes that are associated with one another, pathways, etc), to determine statistical significance between two disease states.

[ivivek_ngs]

Infact , I was quite surprised to see the OP making GSEA a DEG tool which is not. It is important to understand what is the importance of each tools with respect to biological inferences they remit. As @andrew.j.skelton73 highlighted it determines that statistical significance of 2 disease states using information of genes involved in specific pathways.

[Shamim Sarhadi]

Yes you are right, but here http://www.broadinstitute.org/cancer/software/genepattern/gene-expression-analysis , genePattern offers GSEA for DEG analysis , when I tried with that,it collapsed my prob-id and then after t-test and multitest correction ranked them as DEG between case and control !!!

so according, andrew.j.skelton73 and your Ideas, first, I should define my informative genes with limma and then subset them from total data and the next step is using GSEA !!

Thank you

[ivivek_ngs]

If am not wrong the link says does not define them as DEGs if you see here under the pathway section it is using the DEGs to define between 2 biological states the differences of defined state of genes and how they are enriched and in doing that it is performing a t-test for the significance of the state. So that is not a differential expression per se.

Gene Set Enrichment Analysis (GSEA) determines whether an a priori defined set of genes shows statistically significant, concordant differences between two biological states (e.g. phenotypes). The GSEA software packages the method, making it easy to run the analysis and review the results.

Note : If you think my answer is helpful please accept it as an answer so that it can be helpful for others in the community. Thanks

[Shamim Sarhadi]

Thank you vchris_ngs, your answer was completely helpful , I am not vary familiar with biostars structure, I upvote your answer, how can I accept an answer ? I apologize for my noob question !!

[ivivek_ngs]

I believe you can scroll to the place near the up vote for the answer I made and see a tick sign for accepting it as an answer. For the OP it should be an answer.

[GenoMax]

@vchris_ngs: You could repost your comment above as an answer (or use the moderate button and move the comment to an answer) and then @Shamim can accept it. Better that way.

[ivivek_ngs]

Am not speaking about the above comment. The answer which I already gave is put as an upvote and if its of help as an answer @Shamim should be able to accept it as an answer already.

[GenoMax]

Ah I see. You have a separate answer at the top (missed that originally). @Shamim you should accept all answers that were useful using the check mark against the answers (more than one answer can be accepted).

[ivivek_ngs]

Anyone of the the answer mine or andrew can be accepted as answer. We both did answer and are quite detailed according to OP so the OP can accept anyone for community holders. Again am not asking to accept my comment as answer. I see to one of my answer is given upvotes I was talking about that. But in any case since it is having upvote even the answer of andrew is insightful and can be accepted by OP

[GenoMax]

Absolutely. Did not mean to imply that only your answer should be accepted (I will modify my post above). More than one answer can be "accepted".

[ivivek_ngs]

That is what I was intending to say. :)