Ontology meeting 2015-11-12

From GO Wiki
Revision as of 09:39, 17 November 2015 by Paola (talk | contribs) (→‎Refresher on protein binding)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Attendees:Heiko, DavidOS, Harold, Melanie, Paola, PaulT, Tanya, DavidH

Minutes:DavidH


Release Failures

I removed redundant relations from the ontology again this morning after the release failure last night. They seem to be repopulating automatically with an import (UBERON?). If I am reading it correctly, the relations have the same meaning but are represented differently and thus the conflict. Does anyone know why this is happening?

 This is happening automatically. We need to fix it at the source. What should we do for now? For now we should just continue to
 edit. Chris will work on this.

Additional taxon constraints are now being inferred from Uberon

There was an issue because UBERON synonyms with extended characters were creating fatal errors and preventing us to commit taxon rules in OBO-Edit. Now fixed (see https://github.com/obophenotype/uberon/issues/1173) by changing the module extraction pipeline to not include synonyms (or definitions, for good measure) for the uberon import module. Chris, when you're back, could you please add the definitions back? Useful to have them there if possible.

Newly added constraints do not have PMIDs (or other types of references) associated with them. UBERON is usually quite rich in this kind of information. Would be possible to inherit PMIDs from UBERON constraints? Not a super-high priority, but useful - and in line with our approach to try to add references to rules whenever possible.

Also, would it be possible to have an idea of how many annotations across all MODs have been flagged as wrong following the change?

There may still be issues with this. Try clicking on odontogenesis in obo-edit, or search on a child term and have a look at the graph. There are lots of unknown self-referential (circular) relationships.

 Post-poned.

How to import new relations from RO

Can the fake import route be used?

Is it sufficient just to add a URI, use the relation once and wait for the update to run overnight?

 Heiko: Needs to be done by hand in OBO file.
 DOS: Can they come in via gorel?
 AI: DOS to try adding directly_activiates, directly_inhibits to gorel.  
 See if they turn up in _write.obo.  If not, add manually to _write.obo

These should be added manually. It is needed for directly activates and directly inhibits.

Refresher on protein binding

We have a TG FF request for 'protein serine/threonine kinase binding' (as a child of 'protein kinase binding'). Can the group please remind everyone what we agreed on last - is it correct that we are *not* accepting protein family binding terms, but are still fine with simpler protein binding ones? Where did we document this, I [Paola] looked at ontology calls minutes from the GOC meeting onwards and couldn't find it.

 If the idea is to rearrange these terms in some way, then what should we do with these types of terms. Request that they use
 protein binding and column 16. [Paola - DONE]
 Update: True, we are *not* accepting 'X family protein binding' terms; there is a note at the top of TG FF reminding about this.

Missing entries in RHEA or EC to reference GO terms

When we receive requests for GO terms for enzymatic activities that are not yet represented in RHEA or EC, where should we send the RHEA request? Their old SF tracker looks quite empty: https://sourceforge.net/p/rhea-ebi/reaction-requests-updates/

For everyone's reference, this is the current situation: 
Maria Martin at EBI is now leading the enyzme-related resources previously managed by Christoph Steinbeck; 
Thierry Lombardot develops Rhea on the SIB side. 
Thierry confirmed: 
"Yes, this sourceforge tracker/ticket system is still active for Rhea. The URL is correct. Please post your request there and my colleague Kristian Axelson will take care of it.".

Paul T.'s suggested Term naming edit:

1. transcription factor activity, sequence-specific DNA binding -> transcription factor activity

  • Ditto subclasses

2. transcription factor activity, protein binding -> transcription regulator activity, protein binding

  • Ditto for all children apart from cofactors. e.g. transcription factor activity, RNA polymerase I binding -> transcription regulator activity, RNA polymerase I transcription factor binding.
  • The defs of all terms in the branch that are not under co-factor have the clause "may or may not also interact with the template nucleic acid." So cofactor remains a subclass.
AI: DOS to post this as a GitHub ticket to track further discussion, questions.
UPDATE - ticket here
AI: DOS to prepare bulk edits using same system as for previous bulk name changes in this branch.

Synapse project update

  DOS: Major push on this for the rest of November.  
  Latest VU proposal needs work, and some compromises, to get it into GO.  
  Spending next week preparing most of last week of November in Amsterdam implementing.
  Some of what they want may be too detailed for individual GO terms 
 (at least without those terms becoming too complex). 
  So an extension-based approach might be best. 
  This could be designed such that it uses LEGO patterns. 
  May pave the way for using lego in next round.
 DOS and PT need to discuss progress after work in Amsterdam- before management call on 10th Dec.