Ontology meeting 2014-12-09
Attendees: Paola, Heiko, Midori, Tanya, David H, David OS, Jane, Paul T
AI: Paola to put her bits in. DONE.
Numbers are high again, but SF stats suggest that if we had a couple of week-long SF jamborees per year, we'd be able to keep the numbers stable. (We would still have a backlog, given current resources, but at least it wouldn't spin out of control). Shall we try and plan a jamboree around February - is there money for 1 or 2 people traveling? How about bimonthly SF calls: are we interested in having those regularly, or just ad-hoc? E.g. there's a slot available tomorrow.
Re. SF jamboree: Judy not on call; Paul T says we might be able to use roll-over funding before the end of February; not sure what will happen after end of February. (David H could not do end of Jan - start of Feb, but could do end of Feb on the East Coast.)
AI: Paul will take this to the PIs.
Re. SF bimonthly calls: we will have these when needed; today we discussed a ticket from David H (https://sourceforge.net/p/geneontology/ontology-requests/11291/, related to https://sourceforge.net/p/geneontology/ontology-requests/11348/). We may want to try moving the SF calls to alternate Mondays (when LEGO call is not taking place) as Weds work less well.
AI: Paola to email round to explore this possibility. DONE.
David OS reports. Probably 50 terms that will get experts’ review. May be able to get some funding for a few months - details weren’t fleshed out, but there may be potential. Paul S and Paul T were also at the meeting. Participants were keen on doing GO annotation and were interested in GO slims.
Moving relations from RO to gorel
What's the process for this? I know gorel imports a subset of RO, how do we expand that subset? We need the has_symbiont etc relations in gorel.
DOS: Chris can confirm, but I believe the way to do it is to make them subProperties of 'go annotation extension relation' (GOREL_0000000). Will try that today.
Chris: Correct. Technically any logical axiom will do for seeding the module extraction, but subproperties of 'go annotation extension relation' is a good convention for viewing easily in Protege
AI: David OS will look into this.
Enzyme binding - punted to next week
We agreed to obsolete 'DNA methyltransferase binding' from the TG queue a couple of weeks ago, we need to follow up with a plan for how to handle the existing terms in this branch, and guidance as to whether annotators request these terms. They've asked for us to report back on an annotation call.
We took a look, but resolved that we need Chris and Harold on the call to figure this one out properly. So we’ll punt this.
Axiomatizing multi-organism processes - punted to next week
We've agreed it's a good idea to try and axiomatize at least some of these terms before I leave. We need to agree which relationship I should use - can I go ahead and use regulates or do we need the more specific regulates_in_other_species?
Follow-ups from last call - punted to next week
Everyone please take a look at AIs here: http://wiki.geneontology.org/index.php/Ontology_meeting_2014-11-20
Follow-up from last week as Chris and David OS were away: 'Response to' terms again
From this week's assert report:
ADD GO:0097044 'histone H3-K56 acetylation in response to DNA damage' GO:0043200 'response to amino acid'
This seems to come from GO:0097044 being descendant of 'response to stimulus' and having an amino acid as input... same as the logical def of 'response to amino acid'.
This is an interesting case. Is the inference too broad for our liking, or are we happy to keep it? If the former, how do we restrict the ancestor?
Should not be 'response to amino acid.' Wrong genus is asserted for one of the terms between 'response to amino acid' and 'GO:0097044', fix that. The term 'peptidyl-amino acid modification' needs to be sorted out, has_input to 'amino acid residue.' Action item: Jane will consult with Gareth @ ChEBI about which ChEBI term to use for the logical def. What is the difference between 'peptidyl-amino acid' and 'amino-acid residue'? In common usage, these would be the same. What was ChEBI's intention? Ex. ubiquitinated proteins vs. 'normal' protein chain (alpha peptide chain)? What about amino acid derivative vs. modified amino acid? Sort out implications for GO.
Problem is broader. Has_input as a relationship for 'response to x' results in the inability to specify 'what' is being responded to. In this example, we'd expect 'damaged DNA' as the input. ChEBI is not likely to have 'damaged DNA' as an entry. Need a different relationship, 'responds to' perhaps. Action Item (not Amelia Ireland): Chris and DOS to comment on 'responds to'
Chris emails: "Usual issue. Either: (1) make response and metabolic response disjoint (2) Make a distinct relation for every possible genus I favor (1)"
For response-to, regardless of the relation used, I think it would be better to treat the response as the entire process of response, making the relationship between metabolism etc and response to be part-of-some, not SubClassOf. Think in terms of lego, see the cellular response to UV here: http://geneontology.org//experimental/lego/docs/PThomaslego-Whitepaper-2010-03.pdf And from a standard ontology design pattern perspective, it's better to combine disjoint hierarchies rather than isa-overload.
AI: We need to hear from the ChEBI folks about this (Jane was going to speak to them).
AI: David OS will try fixing by changing is_a to part_of.
Follow-up: Protein oligomerization - punted to next week
See discussion from last week: http://wiki.geneontology.org/index.php/Ontology_meeting_2014-12-02#Protein_oligomerization
We couldn't complete this discussion: "New terms: 'protein heterodimer' and 'protein homodimer' not in GO but PRO? Pragmatically, this might be the best solution."
(Paola's action item will be discussed next week.)
Protege 4 vs 5
Chris writes: "Protege 4 is stuck on an old version of the OWLAPI. This is not ideal - we have to do conversion separately and not natively. It's a bigger issue for ontologies like CL that are edited in OWL - lagging behind with the owlapi will cause even bigger diffs as TG edits alternate with Protege edits. From the perspective of software development, it would be hugely advantageous to switch to P5. We can build this from maven easily, and make versions that are in sync with TG. Also, P4 will presumably eventually rot. However, P5 is apparently currently unusable. Can we collect some of the requirements in a doc, place a request on the protege feedback list, and then evaluate our strategy based on the response? Has anyone else rather than DavidOS tried it out? "
Jane replies: "I was using P5 for a while, it was fine for me. It didn't cause any problems with the round-trip. The 'improved' search isn't perfect, but Matt gave us some tips on using it when he was here. I'd be happy to switch over."
Anyone else? Let's collect requirements.
Midori has been using P5 with the phenotype ontology a bit. Search function is lacking. Ideally searches could be performed in a similar way to OE, but are currently very limited. David OS suggests trying out the annotation plugin that’s been around for some time.
AI: David OS will look further into this to determine if we can realistically use P5 short-term and what we’d need.