Ontology meeting 2015-07-23
Attendees:
Minutes: David OS
GO term request with composed uberon terms
[Chris and David OS] Stems from https://github.com/geneontology/termgenie/issues/71
See comments on ticket
Follow-up: Intersections vs. relationships in children of 'protein complex'
Background: to avoid mis-classification of protein complexes that may have the same function or be involved in the same process, we need to retrospectively check all children of 'protein complex' that have intersections (not relationships!) with capable_of function links and capable_of_part_of process links. David OS prepared a list:
list of protein complexes with EC def + axiom linking to MF/BP
HJD is continuing to look at them. Do we want to even create relations between complexes and functions or processes? Why not just do it as annotations? Should be discussed on an annotation call or at the GOC meeting.
Note, there is already a working group on protein complex annotation-related issues, but no editors are involved - maybe DOS could liaise with Birgit and Rama?
standardise way to reference protein in GO axioms
There are a number of ways we can refer to proteins:
PRO has:
/ BFO:0000040 ! material entity is_a BFO:0000030 ! object is_a CHEBI:23367 ! molecular entity is_a PR:000018263 ! amino acid chain [SYNONYM: "peptide" (narrow)] [SYNONYM: "polypeptide" (narrow)] is_a PR:000000001 ! protein ***
ChEBI has:
information biomacromolecule . protein . . simple protein
organic amino compound . peptide . . polypeptide . . . protein polypeptide chain
Pro 'protein' has the advantage that it is the parent class to many specific protein terms we use in the ontology. ChEBI 'protein polypeptide chain' has the advantage of having a chemical classification (an example of why this is useful). I'm not sure what ChEBI 'protein' is useful for.
1. We clear need a term with both characteristics. Could we get Pro to map their protein term to protein polypeptide chain, or would that be incorrect? Could we add some local axiomatisation for GO? 2. We neeed OBO foundry uniquenames for ChEBI protein and Pro Protein in order to avoid confusion whilst editing in Protege.
AI: Make a Pro ticket for this. Recommend they use protein polypeptide chain from ChEBI. They currently just map pro amino acid chain to ChEBI molecular entity. DONE: https://sourceforge.net/p/pro-obo/term-requests/99/
Implement MF patterns using directly regulation relations?
Chris has added relations such as 'direcly activates' and 'directly inhibits' to RO. These have lots of potentially useful applications, for example in definining enzyme activators and inhibitors
https://sourceforge.net/p/geneontology/ontology-requests/11733/
Background and details of relations: An OWL Model of Biological Regulation
Are these ready to use for design patterns for MF, e.g.
ubiquitin ligase activator activity EquivalentTo: 'enzyme activator activity' that 'directly activates' some 'ubiquitin protein ligase activity'
(Note - Does sound a little odd. From its name, 'activates' feels like it should have a continuant as its range...)
Is the def tight enough?: 'directly activates': "p directly activates q if and only if p is immediately upstream of q and p is the realization of a function to increase the rate or activity of q."
David H has been pushing for direct regulation to be reg of an MF or of some MF that is part of a process. This sounds like it could be broader, although maybe that hinges on what 'immediately upstream of' means. Also, in the activator/inhibitor branch, we have:
- enzyme activator activity: "Binds to and increases the activity of an enzyme."
- enzyme inhibitor activity: "Binds to and stops, prevents or reduces the activity of an enzyme."
Does direct_activation/inhibition of MF always involve binding to the agent (GP or complex?)
We also have:
"protein kinase activator activity" is defined as positive reg requiring binding, but the very similarly name 'activator of protein kinase activity' is defined as: "Any process that initiates the activity of an inactive protein kinase."
Do we care about distinguishing activation from zero activty from binding + reg of MF? Do we really need MF and BP terms here?
AI: Make 'directly positively regulates' as parent term for directly activiates (MF to MF only) ditto inhibits. :CHRIS: AI: Work on reconciled definition for direct regulation. Can we abstract the definition sufficiently for RO (not referring directly to MF)? Yes. Chris will add bit to RO def about 1 agent per process and 2 agents interacting.
Jira tickets
Time for our periodic review of Jira tickets, especially unassigned ones. Please go through JIRA issues in advance of meeting. We will aim to devote about 20' of the call to discuss Jira tickets.