COG Annotation - Dealing with genes assigned to two or more COG categories

Question

I am dealing with a list of genes which have been selected from a gene enrichment analysis. In order to see what kind of genes are overrepresented, I ran eggnog-mapper to do an orthology assignment of each gene against a bacterial database (provided in eggnog).

After running eggnog you get a .csv file containing for each gene which is the closest ortholog, and a list of GO, KEGG and COG terms associated.

From there, I would like to do a Fisher-test (also considering Bonferroni correction) to highlight COG categories overrepresented in our dataset compared to our reference genome.

Finally my question: How do you deal with genes assigned to multiple COG categories (same would apply for GO terms). You count +1 to each of the COG categories assigned? Do you just choose one category per gene?

Answer 1

You add 1 to the count of each COG category. You are looking for over represented COG categories so you must count them all. Many genes will be assigned to multiple categories. In fact, a majority of human genes are given multiple GO cellular function annotations. I have worked considerably in this field and one of the surprises I found is how common protein multifunctionality is. But I digress.

The point here is that you want to see whether any COG categories are over represented in your test set. Whether a gene is annotated to one or to a hundred categories is irrelevant. You are counting genes per COG, so you need to count each of the COGs assigned to a gene separately.