Difference GO terms Uniprot and PANTHER

Question

[padgaus]

Hello,

I'm using PANTHER to get enriched GO terms in my protein sets for further analysis.

During one enrichment test i noticed something strange: the protein RBBP4 has the term GO "mRNA splicing, via spliceosome" associated.

So i found that strange (because RBBP4 had other terms already and pretty different than that), i looked up its Uniprot page and i found that Uniprot has different terms associated (i checked with GeneOntology and it agrees with the Uniprot page).

http://www.pantherdb.org/genes/gene.do?acc=HUMAN|HGNC=9887|UniProtKB=Q09028

http://www.uniprot.org/uniprot/Q09028

I emailed PANTHER last friday and didn't get an answer yet. Do you know what could have caused this?

Thanks,

Yujin

Answer 1

[Jean-Karim Heriche]

Different resources have different annotation procedures. Even if they don't do annotations themselves, differences can be caused due to ID mapping and version differences. In your particular case, notice that Panther is using the GO slim ontology and UniProt the full ontology.

[padgaus]

Yes, i understand that but GO slims are trimmed versions of the full ontology so shouldn't the full ontology present all the terms in the GO slim? Or am I missing something?

[Jean-Karim Heriche]

Panther uses its own GO slim to do its own annotations. From this paper:

the GO Phylogenetic annotations are used to annotate PANTHER subfamilies and families. Thus they are added to the previously curated PANTHER (sub)family GO annotation sets (referred to as ‘PANTHER GO-slim’ sets on the website). As a result, the PANTHER GO-slim annotations include all inferred annotations from the GO Phylogenetic Annotation project

[padgaus]

What should i do in that case? They still are different results

[Jean-Karim Heriche]

Different annotations, different results. There's nothing wrong here. What you should do is up to you. You could report everything, making sure you indicate the different sources. What I think is best is to work with one well annotated reference genome and stick to it, i.e. avoid any mix and match between different resources that have different notions of what a gene is or even none at all. As a source of well annotated reference genome, I recommend using EnsEMBL. EnsEMBL integrates various annotation type, including GO terms.