Ontology meeting 2014-06-19
Attendees:Jane, David H, David OS, Heiko, Jane, Tanya, Paul T
Minutes:David H
CHEBI
(Stemming from the 'secondary metabolite' merge issue) Jane to report from the CHEBI workshop.
CHEBI will not classify things a "x metabolite". If we are happy to keep it as a category, then we can maintain it. The thing is it is very species-specific. There is no pressing reason to remove it right now. Do we need it for LEGO? We don't need it for LEGO. Can we define it using disjoint classes? There are certain types of compounds that are always going to be considered secondary metabolites. Keeping it in some union/disjoint way would mean that we will be maintaining the chemical hierarchy for these. Can we define primary/secondary metabolism? The things that are necessary for life are primary metabolic processes. Is it worth the effort maintaining it? Jane will send an e-mail to GO-friends and will ask if this classification is useful. Paul will look and see how we stand in the current ontology. Can we make a union-list for the primary metabolic processes?
Prefixing 'obsolete' to all obsolete classes
Done?
This is to fix a bug that renders the labels by string-matching. David OS will look to see if this is still an issue in Protege5. If we are going to go ahead, Chris will do this.
Change relation for metabolism terms
https://sourceforge.net/p/geneontology/ontology-requests/10881/
We will create more specific relations than has_participant for metabolic processes.
Links to annotation guidance
For background, see http://wiki.geneontology.org/index.php/Ontology_meeting_2014-05-08#Links_to_annotation_guidance
Chris added 'curator guidance link' RO_0002484.
To discuss today:
- how should we record the propagation rule (I,P,R) - axiom annotation
This becomes a bit of an issue when we are concerned with the viral terms. For component terms, can we make a rule that identifies them as part_of virion? Jane will get a list of the examples of the problematic terms.
- using axioms rather than regexps
- consistent location of docs: go wiki, drupal, or go nuts?
Chris had commented by email: "Propagation of annotation guidance - open to ideas about how best to encode the propagate-down rules. Worst case we could just add the UniProt codes and regexps as a property but I'd rather than a more logic-based way to do this."
'negative regulation of response' terms
Stemming from this SF ticket - could you all please take a look and comment:
https://sourceforge.net/p/geneontology/ontology-requests/10927/
Delay until next week.
Should there be limits to generation of compounds of compounds using TermGenie?
We've recently had some very long terms requested in TermGenie. e.g. 'negative regulation of protein localization to cell tip involved in positive regulation of establishment of cell polarity regulating cell shape'
I'm reminded of the definition of sub-editing favoured by the chief sub at Nature: The readers last line of defence. Is there a last line of defence for GO users?
These terms are still in the TG queue pending discussion. See email threads:
- Re: [go-ontology] Stand back and admire the combinatorial explosion ?
- Re: protein localization to cell tip involved in positive regulation of establishment of cell polarity regulating cell shape
Immediate issue: can we either:
- Find a way to simplify these terms so that a regular human could follow them
- Suggest an equally expressive alternative using annotation extensions
- Suggest that Val record something simpler (or are there no limits to what we will let annotators try to cram into TG + AE in the absence of LEGO?)
Poss solution: Are we over-using regulation for cases where one process causally depends on another? If so we may be able to strip out most of the regulation here. Perhaps:
protein localization to cell tip involved in establishment of cell polarity (regulating cell shape)
Bigger issues:
- Should we try to come up with some rules to limit compounding of compound terms?
- Is more expressiveness always a good thing? Surely not if it makes annotation unreadable for most humans...
Do we just table this and allow LEGO to arrive? For what are these types of terms useful? Can we flag terms that are no longer allowed to be used in a new compound term? We should document when to use regulation terms.
AI: Review and potentially revise doc on regulation. AI: Make proposals for limiting the generation of compound terms in TermGenie - triggering a warning asking for Ticket submission of some critical level of compounding hit.
Using Tawny to retrofit metabolism and transport to fit new patterns
{Added this in case Chris turns up with Phil}