Question: Enrichment Analysis in R
gravatar for Bayram Sarilmaz
10 weeks ago by
Bayram Sarilmaz0 wrote:

I have two gene sets: A and B. I would like to check which genes in B are enriched in A. As a result of the enrichment analysis, I to have a p-value for each gene in B.

Here is a reproducible example that you can use: I'm performing the analysis of gene IDs.

A = data.frame(c(100,200,300,400,500,600,700,100,800,900,1000,100,500,100)) #Gene IDs in set A
B = data.frame(c(200,4,900,100,6)) #Gene IDs in set B

#check if B geneIDs are enriched in set A, and generate a p-value for the enrichment of each gene
go.obj <- newGeneOverlap(B,A)
go.obj <- testGeneOverlap(go.obj)

In the above example I attempted using GeneOverlap package, but it didn't give me p-values for every gene in B. Any suggestions on other methods to achieve what I'm aiming for?

enrichment genetics R • 219 views
ADD COMMENTlink written 10 weeks ago by Bayram Sarilmaz0

You cannot generate per gene p-values for enrichment, only a "set" level enrichment, i.e. does set B overlap with set A more than expected? Maybe you could elaborate on your question/goal? To me, "I would like to check which genes in B are enriched in A", means which genes in B are also in A....

ADD REPLYlink written 10 weeks ago by ejm32420

Thanks for explaining this! Then if the p-value from the above geneoverlap test is equal 0, it means there is no significant overlap between A and B?

ADD REPLYlink written 10 weeks ago by Bayram Sarilmaz0

You should question P values that are equal to zero, particularly when you're dealing with just 5 elements in your B object and when a visual inspection reveals that only three-fifths of B form a subset of A. Why not just report that, i.e., that 60% of B overlaps with A? Why do you need a P value when human eyes are sufficient? Presumably your actual gene lists are much larger?

ADD REPLYlink modified 10 weeks ago • written 10 weeks ago by Kevin Blighe30k

See @Kevin's comment below about meaningfulness of your p-value. But to your direct question, No a p-value of 0 would mean that the overlap is statistically significant. I.e. there is a 0% chance that you would find this degree of overlap or greater if there was no association between the two lists.

ADD REPLYlink written 10 weeks ago by ejm32420
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1195 users visited in the last hour