GO annotation when the paper gives no biochemical evidence for molecular function?

[martin2urban]

Hi @Hsinyugithub , a quick GO convention question from curating PMID:42089373 (the F. pseudograminearum Sdh subunits).

Their molecular function is textbook (succinate dehydrogenase activity, GO:0000104) and already in UniProt, but this paper does no biochemistry — all the experimental evidence is genetic (deletion mutants losing growth, conidiation and DON).

So my question is: when a paper provides no biochemical evidence for a gene's molecular function, do we still annotate the MF term — and if so, with what evidence code, given ISS is out and TAS is our fallback — or do we keep GO to what the deletions actually demonstrate?

I believe, we assign the GO term annotated by UniprotKB or perhaps this is not correct? Thanks.

[Hsinyugithub]

We could use "TAS" in this case. UniProt is a good source for GO annotations; however, many of the terms in UniProt are based on predictions. We will try to avoid using terms that lack experimental evidence, unless the function is explicitly mentioned by the authors and is important to the overall experimental design of the paper.

I think another evidence code that might be useful for many of our articles is "Inferred from Sequence Orthology (ISO)". But I remmeber our team has truned down the usage of this code previously.

[martin2urban]

Thanks @Hsinyugithub — that settles it. Recording the rule as:

TAS is allowed when the authors explicitly state the function and it matters to the experimental design. The justification is the authors' statement in the paper, not the fact that UniProt carries the term — which rules out bulk transfer of UniProt GO annotations, since those are largely pipeline predictions with no author behind them. The emphasis stays on experimentally verified annotations, with TAS as a narrow, deliberate exception.

For PMID:42089373 that gives GO:0000104 by TAS (succinate dehydrogenase activity is the premise of the study, not an aside), alongside the process terms the deletions actually demonstrate — growth, conidiation, DON production — by IMP.

I'll add this next to the existing ISS/TAS rules in phiweaver's conventions doc.