issues with the hypergeometric test for gene enrichment analysis

Question

[Marouen Ben Guebila]

Hello everyone, I am trying to perform gene enrichment analysis of a set of genes. The test I have used in matlab is the hypergeometric probability density function. The issue is that I noticed that some pathways had a significant p-values although they had 0 genes counted. Thanks for your help.

[Devon Ryan]

Are you doing a two-sided test? You likely don't want to do that.

[Marouen Ben Guebila]

Actually, I thought about that, though the matlab function hygepdf, nor the r function dhyper have the argument for one sided test. https://stat.ethz.ch/R-manual/R-devel/library/stats/html/Hypergeometric.html

[Devon Ryan]

I assume that there's an equivalent of the R fisher.test function in matlab. Try using that and seeing if the results differ. Fisher's test is the same thing but will involve less reinvention of the wheel.

[Marouen Ben Guebila]

Devon, maybe you can help with the following situation I have a list of upregulated genes, and I've downlaoded GO pathways in my system. How can I proceed? I am aware of the tools available out there, I am not intending to reinvent the wheel it is just part of a pipeline. thanks!

[Devon Ryan]

As WouterDeCoster aptly wrote, don't reinvent the wheel. There are many many packages already existing to do GO analysis, just use one. You'll likely need to stop using matlab for this, but that's no loss (it's rather uncommonly used in bioinformatics).

[WouterDeCoster]

I think there are many tools you can just plug in to your pipeline, rather than as you say reinvent the wheel. (e.g. topGO in R)

Answer 1

[EagleEye]

Use this simple tool, no installation required, easy to integrate as part of any pipeline,

Gene Set Clustering based on Functional annotation (GeneSCF)