Ontology meeting 2016-11-10

From GO Wiki
Revision as of 13:40, 10 November 2016 by David os (talk | contribs)
Jump to navigation Jump to search

Attendees:

Minutes: David OS?

Regrets:

GoToMeeting invite: https://global.gotomeeting.com/join/859015101

GitHub projects page (we'll try to follow this, as a test - see columns "Ontology meeting TBD" and "Ready to be worked on"): https://github.com/geneontology/go-ontology/projects/1

Agenda below is there as backup.

Debrief from GOC meeting: ontology group update and breakout session on ontology priorities

For reference, minutes are here: https://docs.google.com/document/d/1MAnnOfs-e2LY9MnqdCZscalbxbNUDSJ9pbMZ-f2WS9U/edit#

Discussion of 'positive regulation' and 'maintenance' from last annotation call

See http://wiki.geneontology.org/index.php/Annotation_Conf._Call_2016-10-25#Discussion_of_Outstanding_github_Tickets_2

Modified Proteins Proposal

The proposal was not well received by annotators at the GO meeting. The annotators thought that these terms were too important biologically for them not to be in the ontology. We need to come up with alternatives.


  1. Continue to add these terms by hand. Unsustainable?
  2. Figure out a way to add these terms automatically. Use PRO?, Chebi?, PsiMod?, will this lead to annotation inconsistency. Probably.
  3. Only add a few very high-level terms. What is the specificity cut off and what happens when annotators make a case that the very specific term that they want is important.
  4. Have do-not-annotate' terms that are generated automatically as in #1 below and categorize gene products by the entity that is captured in binding annotations.

We discussed three possible semantic interpretations of modified protein binding terms:

https://docs.google.com/document/d/1MAnnOfs-e2LY9MnqdCZscalbxbNUDSJ9pbMZ-f2WS9U/edit#heading=h.vzsyf1ss0k8h

1. Binds a protein that happens to have a modification, but not necessarily the modification site <- Editor's don't want this. No-one else does either. 2. Binds to the modified part of a protein <- Sylvan wants this. Seems to be most broadly supported. 3. Binding to some specific protein partner is dependent on the modification state of the partner even if binding is not to the modified bit. <- More controversial Pascale likes this, some other seem to support. DOS objection: this seems like annotating a property of the bound protein. Also specific to one binding interaction (whereas 2 is likely to be a general function)

Number 2 in this list - binds to the modified part of a protein - had the most support. Example: SH2 domain confers binding to phospho-tryosine in the context of a specific peptide motif: https://en.wikipedia.org/wiki/SH2_domain

For terms like this, we could add a comment that it should only be used where there is high confidence that binding is to the modification + protein (simple dependency on phosphorylation of target is not sufficient but one that localizes the domain by deletion/mutagenesis analysis of the protein is). This still leaves the question of how detailed we should get in specifying the target (see histone binding).

Follow-up: contacts for UniProt annotations

Background: http://wiki.geneontology.org/index.php/Ontology_meeting_2016-10-13#Update:_contacts_for_UniProt_annotations

AI from last call was "Sylvain will discuss with Chris and David H at the GOC meeting". Any news on this please?