I have an RNA-seq data that I have analysed using Limma-voom and have extracted the gene IDs, log2FC and the p-values. At p value < 0.05, I have over 10,000 DEGs, however, when I run the GO analysis, I hardly get any GO enrichment terms at the set pvalue_cutoff of 0.05 because the pvalues for the GO analysis are almost tending to 1.00. Is it possible to have such number of DEG genes and hardly get any GO terms? If I should relax the cutoff for the pvalue, what acceptable pvalue above 0.05 won't be too ridiculous or that will be acceptable?
Thank you!
Hi.
First, you should select your DEGs based on adjusted p value (NOT p.value). Furthermore, 10k DEGs seems too much! You should set a logFC threshold (e.g. |logFC| > 2) for the selection of DEGs to have more reasonable DEG numbers and run GO on those.
Hello Olatunde Omotoso,
yes, it could be possible to find such huge numbers of significant differentially expressed genes (DEGs). It depends on which two groups you are comparing. Nevertheless this is not very valid, if using only p-values or even corrected p-values. In chip data analyses we used a multi-parameter method (High-Performance Chip Data Analysis, HPCDA), that includes also p-values but a lot of other relevant data, as could be seen in my profile and the added links. You should analyze your RNA-Seq data in a similar way and then upload the resulting DEGs into GO. Surprisingly a fold change (FC) was not advantageous to get the list of significant DEGs with HPCDA. Using a FC of 4 (log2 FC of 2) could loose a lot of relevant genes, but this depends on the number of individual datasets in both groups.
https://www.researchgate.net/profile/Joachim-Gruen
Method Effect size calculations as impressive method to differentia...
Technical Report Towards a more quantitative view of gene expression profilin...
- Badges
- Science topic
- Similar topics
- Biological Science
- Biochemistry
- Biochemical Techniques
- Analytical Chemistry
- pH