RNAseq, GO enrichment, and length bias correction

[lukas1848]

Hi,

I have a set of DE genes from an RNAseq study in a non-model species. I performed a GO enrichment analyses using topGO and GOseq.
TopGO is nice, because it offers the parent-child algorithm, which takes into account the hierachical relationship between GOs. GOseq is nice, because it corrects for length bias in RNAseq.
Yet, of course I would like to combine both advantages. Ideally, I would just run topGO and perform a length bias correction on the calculated p values.
I found a paper by Gao et al. (http://bioinformatics.oxfordjournals.../27/5/662.long), where they essentially describe how to do this, at least that is what I understood, but to be fair, I am neither a bioinformatician nor a statistician.
The problem is, the R scripts mentioned in the paper are no longer available.

So my question is: Is there a simple way to correct for the RNAseq length bias, by just correcting the p values returned by topGO's fisher tests? I imagine to just multiply the p values by a correction factor based on the average transcript length in the respective GO term.

I hope some of you might have an idea on that.

Thanks,

Lukas

[chadn737]

Did you try emailing the authors about the scripts?

[lukas1848]

yes, I did. The author will try to fix the link. But I am not sure that I will find a simple answer to my question in the script.

[lukas1848]

yes, I did. The author will try to fix the link. But I am not sure that I will find a simple answer to my question in the script.

Ok. I got the scripts, but they didn't really help me in answering my question.

Edit: I got another answer from the author and unfortunately, it's not as simple as I hoped. I have to estimate the noncentral parameter for the Wallenius distribution based on transcript length before I calculate my p values. This won't be easy to implement into my topGO workflow. Is anyone here aware of a program that can do both, lenght bias correction and parent-child testing for GO enrichments?