Bioconductor Goseq - Overrepresented P-Values

Question

[alittleboy]

Hi All:

I have a question about the bioconductor goseq package for GO enrichment analysis. Those top-ranked categories are obtained based on the ranking of "overrepresentedpvalues" from the goseq object. The goseq also includes "underrepresentedpvalues" from the same output. Can I know how the over/under-representations are determined?

My question can be probably generalized in this way: can I say if there are more DE genes for a particular category, then this category is "enriched" and the associated p-value is called "over-represented", while if there are fewer DE genes for a particular category, then this category is called "depleted" and "under-represented"? Can this be reflected in the sign (+/-) of certain statistics?

I am new to this area, so thank you very much for your help! The vignette of the goseq package can be found here.

Answer 1

[Chris Evelo]

If I understood the vignette correctly goseq actually does not do any overrepresentation analysis itself.

You can use it to correct for the length bias in RNA-seq data that would otherwise make you overestimate genes, and thus GO classes with those genes, that have longer sequences.

The vignette says:

goseq will work with any method for determining dierential expression and as such dierential expression analysis is outside the scope of this document, but in order to facilitate ease of use, we will make use of the edgeR package to calculate dierentially expressed (DE) genes in all the case studies in this document.

So if you want to understand the case studies described you might want to read the edgeR documentation.

I would also advice you to read the [?]GO_Elite documentation[?]. That is work in progress, but we wrote it to help you understand some intricacies in GO analysis in general.

[Chris Evelo]

Couldn't edit my own (old) post. Wanted to add that a GO-Elite paper has now been published. It is at: http://dx.doi.org/10.1093/bioinformatics/bts366