Ontology meeting 2016-01-28

From GO Wiki
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

Attendees:

David OS, Harold, Heiko, Paola, Tanya, Chris, Melanie


Minutes: David OS


TermGenie review queue

Just a reminder that the following term requests are still in the queue - do they need obsoleting or following up on? Or if more discussion is needed, they might better be moved to GitHub:

cholecystokinin secretion

meiotic APC-fizzy-related complex (Tanya:GH11977 and GH11757 are related, need to take to annotation call)

insulin-like growth factor production

Discussed all this before.  David Hill has been working on the merge but has encountered some issues.
We have a comment for production terms under cytokine production. 
This states that we have them because in some cases we cannot tell what the increase in the cytokine level is due to.
AI: obsolete pending secretion and production terms in TG review queue. DONE
We need guidance somewhere to stop people requesting them.  But where to put this:  
Chris suggests in comment on design pattern in GitHub design pattern site.  DOS TODO
 DONE https://github.com/dosumis/dead_simple_owl_design_patterns/issues/15
 Suggest we use description field.  Comment on ticket if you disagree.
 Chris & DOS discussed making pretty looking GitHub IO site based derived from design patterns.
 


Lys63-specific zinc metallopeptidase deubiquitinase activity (Melanie: this is linked to the GitHub at https://github.com/geneontology/go-ontology/issues/12182. We have resolution for this since this morning, but there is the bigger issue raised by https://github.com/geneontology/go-ontology/issues/12184 - where we discussed that there shouldn't be activity specific of a type of enzyme, but rather just the general activity (e.g. deubiquitinase activity). Antonia couldn't annotate to 'GO:0061578 Name Lys63-specific deubiquitinase activity' due to the parentage issue in 12182. If we decide that we want to remove all activity specific terms such as with phosphatase, we'll need to work on quite a bit of refactoring)

General discussion of how we choose how specific MF terms should be  - vs extended description in LEGO or with AE.   
 We generally want to avoid naming activities for gene products.  But we have lots of cases like this.  Various reasons they could be justified.
 
 Need to collect these and turn them into guidance
 AI: Ticket for Paul Thomas to start writing some guidance. Done. GH12257

Pending Jira issues in 'old' EBI Jira instance

Another gentle reminder :-) that we resolved to move these to the appropriate GitHub trackers. Some may be out-of-date by now, but we shouldn't forget about the others:

https://www.ebi.ac.uk/panda/jira/browse/GO/?selectedTab=com.atlassian.jira.jira-projects-plugin:issues-panel

Related: I (Paola) have a ticket about creating a TG template for 'channel activity' terms. But ever since we discussed this 10 months ago, I don't think there have been requests for such terms. The branch is quite developed already (DOS, I think you worked on it a bit as part of the synapse project?). Seems like there won't be a big need for such a template. Drop task? (https://www.ebi.ac.uk/panda/jira/browse/GO-343)

YES - done

Logical definitions for 'response to chemical' terms

See background here: http://wiki.geneontology.org/index.php/Ontology_meeting_2016-01-21#Logical_definitions_for_.27response_to_chemical.27_terms

Follow-ups: Val agrees; DOS comments: "There are 413 subclasses of 'cellular response to chemical stimulus', so I don't really fancy changing this by hand. Terms in this branch are so consistent, that this could be a candidate for completely automated generation of a branch based on specifying applicable patterns. However, it doesn't feel like a very high priority, given that almost all multiple inheritance in this branch is inherited from chebi and that the current pattern works fine to automate classification based on chemical input."

Shall we (DOS) create a GH ticket for this so it doesn't get forgotten? Or shall we just drop it?

 DOS to create a ticket with the label  or Milestone 'In the fullness of time'
 DONE: https://github.com/geneontology/go-ontology/issues/12255

Merging of terms

There is a proposal to merge some phosphatase terms, see https://github.com/geneontology/go-ontology/issues/12184 For example, "protein phosphatase type 2A inhibitor activity" would be deprecated and become a narrow synonym for "protein serine/threonine phosphatase inhibitor activity"

As we don't want IRIs to disappear, do we just deprecate those terms and point to the one we want to keep as their replacement?

 Plan is outlined here: http://gocwiki.geneontology.org/index.php/Ontology_meeting_2015-07-02
 For now we merge as normal, but obo2owl conversion is meant to be updated (extended with additional job?)
  to make obsolete class in OWL from alt_ids (which stay in palce)
 Chris to make sure only one replaced_by axioms per obsolete term.
  AI: guidance on merge in ed manual needs to have losing synonym get alt_id tag.
  Decided we will retain alt_id (in addition) long term.


CL vs CLO

The term CL:0000002, immortal cell line cell, became obsolete. It has been replaced by http://purl.obolibrary.org/obo/CLO_0000019, as mentioned at http://purl.obolibrary.org/obo/CL_0000002, though it did lose it's definition on the way. Consequently, annotations in which we use that term are now invalid. Not strictly an editor question but more relevant to the ontology side of GO nonetheless. CLO is also not in the GO list of dbxrefs for example.

 AI: should be dropped from annotation extension.  Should aim to switch to ECO code instead.

Last week's inferences

Follow-up to http://wiki.geneontology.org/index.php/Ontology_meeting_2016-01-21#Latest_assertions

ADD GO:0000740 'nuclear membrane fusion' GO:0000741 'karyogamy'

Karyogamy is the fusion of the entire nuclei - membrane plus cytoplasm. So 'karyogamy' has_part 'nuclear membrane fusion'. (https://en.wikipedia.org/wiki/Karyogamy) It's probably safe to say that 'nuclear membrane fusion' is always part_of 'karyogamy' but definitely not an instance of it. I know, we decided that 'organelle membrane fusion' is_a 'organelle fusion'...

 DOS: Quite a bit of useful inference now dependent on this 
  (e.g. various vesicle fusion events important to synapse project).  
  Would have to review effects of changing this 
 Maybe the solution is to treat karyogamy differently - 
  From https://en.wikipedia.org/wiki/Karyogamy - looks like it also includes the fusion of nucleoli.
 So could move out from under organelle fusion.
 BUT - karyogamy has some children named for fusion of nucleoli. How to deal with these if changing parent class?
 AI: Paola to make ticket.  Solution still not clear...
 DONE: https://github.com/geneontology/go-ontology/issues/12256, assigned to DOS due to possible consequences for synapse terms.

Regulation terms should not have regulation children

We discussed this before but I can't find minutes. I *think* we agreed that TG will prevent addition of regulation terms if the parent is itself a child of 'regulation of biological process'. I also think we said this may be more difficult to implement for children of 'regulation of biological quality'. I do not remember if we resolved anything about *existing* terms that already have this double regulation parentage. This has implications for annotation. What prompted this today was a request for 'regulation of vasodilation' terms. Looks like the branch of 'regulation of blood vessel size' was worked on years ago and we may need to correct it. The main point is - if 'vasodilation' is_a 'regulation of blood vessel size' and is defined as "An *increase* in the internal diameter of blood vessels, especially arterioles or capillaries, usually resulting in a decrease in blood pressure." (which is great), then why do we need 'regulation of vasodilation' ("Any process that modulates the frequency, rate or extent of increases in the diameter of blood vessels.")? Regardless of how often, frequently or prolongedly a protein increases blood vessel size... it's still increasing it. Looking at the abundance of children and descendants, the situation is very convoluted. Same for 'vasoconstriction'. Additionally, both vasodilation and vasoconstrictions are is_a 'biological regulation', while they should be more specifically is_a 'regulation of biological quality'.

For reference, the GH ticket is https://github.com/geneontology/go-ontology/issues/12253

Discussed briefly. Agreed that work needs to be done. Details will be posted on the GH ticket as we go.