Question: what is the cut-off for saying hat a gene set is positively/negatively represented in GSEA?
0
gravatar for rhasanvandj
15 days ago by
rhasanvandj0 wrote:

what is the cut-off for saying hat a gene set is positively/negatively represented in GSEA? I have 2 phenotypes and I have done GSEA. Based on FDR <0.25 no gene sets are significantly represented in the GSEA output. Why is that? Is FDR<0.25 the correct cut-off for GSEA? I have a large gene data and some of them were significantly differentially expressed. but why my GSEA have weird results?

rna-seq software error • 155 views
ADD COMMENTlink modified 14 days ago • written 15 days ago by rhasanvandj0

Which implementation of GSEA are you using (there are many)? - please show code, input, and output, where possible. Your definition of 'weird results' is neither well defined; so, please define it further. I would be skeptical about using 25% FDR.

ADD REPLYlink written 15 days ago by Kevin Blighe61k

Hi Kevin Thanks for responding. I dont know excatly what to tell yu. I am using hallmark in GSEA. What do you mean by code? I am running GSEA in server for broad institute. input is gct file and cls file for 2 phenotypes (patients having feature=1 and no feature=0) in the output no gene set is significantly regulated based on FDR<0.25 cut off. this cut off is in the guideline of GSEA server.So, what cut-off would you use? What else do you want me to tell about my analysis?

ADD REPLYlink modified 14 days ago • written 15 days ago by rhasanvandj0

Please use ADD REPLY to comment. Given you used the tool correctly and there are no significant results then it is what it is. Your data are not generally enriched for the gene sets you tested. Having many DEGs is not any guarantee for a significant GSEA result. GSEA tests if globally your transcriptome in one vs the other condition is enriched for a certain gene set, so if most of these genes tend to have high or low ranks based on your ranking method. You might try other methods such as enrichment profiling of only the significantly upregulated genes with tools such as g:profiler2 using all genes you analyzed as a background.

ADD REPLYlink written 14 days ago by ATpoint36k

As per ATpoint, this may very well be the genuine result. Ensure that your gene names match those of the pathway / signature against which you are enriching, too (i.e. Ensembl IDs == Ensembl IDs). Other than that, as we cannot really see your screen and walk through this for you (and we neither have the time for that), I can only suggest that you contact Broad Institute if you still feel that there is an issue with their program.

ADD REPLYlink modified 14 days ago • written 14 days ago by Kevin Blighe61k

Thanks GSEA only accept alphabetical IDs for genes ( i.e MUC16) and IPA for pathway analysis accept numerical values (i.e.94025 ). Differential gene expression can be done with mixed IDs (i.e. MUC16|94025) I dont think this can be the reason for weird result. what is your idea?

ADD REPLYlink written 14 days ago by rhasanvandj0

Hey, you still have not specifically defined "weird result" (?). Also, the gene IDs should match exactly.

Note: the numerical IDs are Entrez gene IDs.

ADD REPLYlink modified 14 days ago • written 14 days ago by Kevin Blighe61k

Weird result mean no significantly enriched based on FDR <0.25. How should the gene IDs be similar? different programs accept different IDs format.

ADD REPLYlink written 14 days ago by rhasanvandj0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1596 users visited in the last hour