search

geneontology / go-ontology

Source ontology files for the Gene Ontology
http://geneontology.org/page/download-ontology
Creative Commons Attribution 4.0 International
220 stars 40 forks source link

GO term specificity proposal 1 (generic residues and regulatory subunit dependencies in term names) #12416

Closed ValWood closed 8 years ago

ValWood commented 8 years ago

We always try to annotate as specifically as possible, and encourage our community to do the same. This has highlighted a number of pages where annotations are sometimes get modelled arbitrarily slightly differently .

Take for example cdc2 MF annotations below:

specificity

The "protein kinase activity" is described variously as cyclin-dependent protein kinase activity cylin-dependent protein serine/threonine kinase activity protein kianse activity protein serine/threonine kinase activity.

First, cyclin dependency can be captured by annotating the cyclin correctly. Secondly the residues are not useful in this context. At the level of GO annotation, it isn't informative to know which general residues are phosphorylated. Its only useful information if you know which specific residues are phosphorylated in a particular target protein. This should/could be captured on the modified protein, using PRO terms.

e.g http://www.pombase.org/spombe/result/SPAC24B11.06c#modifications

but including this information in the function seems out of scope for GO? (also applies to protein phosphatase activity descendants)

There are various tickets where this has been discussed, but definite no proposal to merge.

@pgaudet @thomaspd @dosumis

ValWood commented 8 years ago

I'll link the older tickets with background to this if I find them....

ValWood commented 8 years ago

similarly peptidyl-serine phosphorylation and peptidyl-threonine phosphorylation could merge to protein phosphorylation

ValWood commented 8 years ago

A similar issue was discussed here with respect to ubiquitin hydrolase residues: https://github.com/geneontology/go-ontology/issues/12182

jimhu-tamu commented 8 years ago

Hi Val, saw this series while coming to post an unrelated issue.

You might also consider the various histone modification terms as things that would be affected by your proposal.

ValWood commented 8 years ago

In previous discussions (still haven't tracked down the ticket, we thought that the histones would be a special case). These terms are more meaningful because they refer to a specific residues where modifications are directly connected to particular biological processes (transcription, repair, checkpoint signalling etc). i.e I think we would still want to represent the processes of the histone code.

dosumis commented 8 years ago

Specificity of kinases for specific types of aa residues (e.g. Ser/Thr vs Tyr) is well within expressiveness that GO should capture and name terms for. I have no major objection to having classes for cyclin dependency of that activity either - especially if users expect to see this as a named class and will find it hard make use of the decomposed version. The names in these cases are not too long and are easy to understand.

ValWood commented 8 years ago

We have had a number of discussion about the residues with @pgaudet Paulthomas (needs tagging github ID) and within Pombase @antonialock

They aren't useful, because they don't tell you anything about the reaction (you don't know the specific residue, so on its own this information doesn't add anything to the function annotation).

I think there was general agreement about this at the LEGO meeting?

tberardini commented 8 years ago

Closing after discussion during editors' call. Terms will be retained, new terms with specific residues will not be created as they have not been in the past.