GO analysis and the universe set of genes

Question

[tiago211287]

I started to analyse GO enrichment with my genes list using the R package clusterProfiler, that get data from DAVID. Then, a doubt came out about what set of genes should I use in my universe when looking for enrichment?

All annotated genes?

Only annotated genes expressed in the tissue I am studying? (Mine is heart mouse)

Only genes present in my GTF file?

Thank you

[Jean-Karim Heriche]

Enrichment analysis is based on comparing two lists and what these two lists should be depends on the question one asks. Which question are you interested in ?
1 - Does my list contain more genes of type X than expressed in the mouse heart ?
2- Does my list contain more genes of type X than found in my GTF file ?
Depending on the context (i.e. what the list and the GTF file represent), both are valid questions.
Typically, the background gene list is made of all genes tested in the experiment.

Answer 1

[Martombo]

filtering for genes that are expressed in your samples is a good idea. for example: if you have a neuroblastoma cell line and you pick randomly some genes that are expressed in it, you'll probably find some enrichment in neurological functions. using as background only genes that are expressed should correct for this bias.

[Martombo]

of course the question of what threshold you use to say that a gene is expressed still remains.