Hi, I have what I think are three simple questions regarding the goana() function of edgeR, which I am not very familiar with:

1-Does it work with Arabidopsis?

2- I understand that if the coefficient you are looking into is cond.B - cond.A, you'd be fishing out terms that are differently represented between B and A, if I am right. What if you want to compare GO terms from DE genes in cond.B - cond.A against the GO terms from the whole genome or from all genes in the dataset?

3-If looking into cond.B - cond.A will render terms that are differently represented between B and A, what if you are looking into an interaction-type contrast such as (cond.A1 - cond.B1) - (cond.A0 - cond.B0). What would be the interpretation of goana results?

Thanks a lot for any help you may provide me with on those.

Best,

David

Does it work with Arabidopsis?

Typing ?goana brings up a help page that lists possible species. The pages refers you to ?alias2Symbol, which gives a comprehensive list. No, Arabidipsis isn't one of them. For more explanation see:

• Using kegga and goana for Plants
• [goana] function within edgeR not working for Arabidopsis

You could however convert the Arabidopsis GO annotation to a data.frame and then use kegga instead of goana to do the same analysis:

• how to use "non-standard" species for KEGG / GO analysis in limma?

I understand that if the coefficient you are looking into is cond.B - cond.A, you'd be fishing out terms that are differently represented between B and A, if I am right.

No, not exactly. goana is looking for terms that a over-represented in the DE genes, not merely associated with the DE genes.

What if you want to compare GO terms from DE genes in cond.B - cond.A against the GO terms from the whole genome or from all genes in the dataset?

That's exactly what is done. That's what we mean by over-representation. See the universe argument.

what if you are looking into an interaction-type contrast such as (cond.A1 - cond.B1) - (cond.A0 - cond.B0). What would be the interpretation of goana results?

goana works the same for any contrast. It looks for GO terms are over-represented amongst the DE genes relative to all annotated genes in the dataset.