I have a non-model organism and I tried to use two different strategies to do functional annotation (GO and KEGG):

1.Run local blastx against the entire nr database, then use blast2go to do GO annotation. For KEGG annotation I used KAAS (KEGG Automatic Annotation Server).

2.Run eggnog-mapper for both GO and KEGG

Here's what I got (total # of genes: 22129)

                    genes blasted with hits       genes with GO          genes with KEGG
blast2go+KAAS       20162                           16868                7023
eggnog              16838                           7086                 7983

Blast2GO took a month to complete while eggnog only took 2 hrs. It seems that I got more GO annotated genes using blast2go but there are only ~3-5 GO terms associated with each gene. For eggnog, I got significantly less GO annotated genes but each gene is associated with ~20 GO terms.

I got similar results for KEGG using both methods.

Which result should I use for the subsequent functional enrichment analysis?