CJM EBI Visit June 2009
From GO Public
Participants: Chris Mungall, Jane Lomax, Jennifer Deegan, Midori Harris.
See the walk-through on the implementation plan wiki page.
Note that the "save the file" procedure will change once we have some xps in the live GO and others still in separate files.
Status overview and release plan
- add progress notes to table on xp category page
- look into whether wiki-fu can automate inclusion
- need better documentation of what's in go_write/go_ext (not just xps)
Internal xp status
Some internal xps will go into go_write (and go_ext) soon. Draft order of incorporation:
- 1. regulates
- no new relationship types
- recapitulate existing links
- D&T need for reasoning, to streamline workflow, etc.
- 2. bp_xp_self part_of xps
- no new rel types
- recap existing links
- 3. cc_xp_self part_of xps
- analogous to bp_xp_self
- 4. cc_xp_go
- 2 or 3 new rel types
- depending on whether we're committed to GO MF is_a BFO process (if so, change all has_function to realizes)
- all new links
- very straightforward set
- 2 or 3 new rel types
David recommends adding in order they're cleaned up (mostly consistent with draft above).
Other internal xps will remain in separate files for the time being
- needs lots of new relationship types (and Alan doesn't like them)
- create go/ontology/xps/ directory; move vetted xp set there; continue to keep unvetted in go/scratch/xps/
- can fast-track ones that use occurs_in - this could be #5, or even #4
- bp_xp_self, using relations other than part_of
- cc_xp_self, using relations other than part_of
Once xps are in live file, we have to run the reasoner frequently.
Must notify all ontology editors, and announce on GO Friends that these will be in go_ext. Also want to give public more of a coherent plan.
- add xp overview to GO Timeline: Midori
- email gofriends: David and/or Midori
- news item: Jane
Chris and Jen worked on the Biological_proces_xp_self set and the work is minuted at Biological_process_xp_self_ProgressNotesNov2008#10th_June_2009
External xp status
- GO-CL: waiting for improvements in Cell ontology [also something else to fill in, about things GO can do in the meantime ... but I didn't get specifics down]
- GO-ChEBI: see below
- challenge is which anatomy ontology(ies) to use
- possibly use UBERON for non-plants
- tiny set of bp-xp-fungal anatomy can stay the same (keep using FAO)
- [something about plants ...]
Additional participants from ChEBI: Paula deMatos, Marcus Ennis, Janna Hastings, Steve Turner
overview of xps, using go-chebi examples. See XP:biological_process_xp_chebi
note that GO would not have to mirror every rung in ChEBI classification "ladder" as long as go and chebi graphs are consistent
ChEBI plurals - not a big problem, but GO will need guidelines/rules for which chebi term to choose
e.g. "indole and derivative metabolism" would point to chebi "indoles"
at present, go curators manually search, read, and link to chebi entries
want to make go's representation of links to chebi more accessible to computation & reasoning - ergo xps
there are probably a lot of go terms that ought to point to different chebi entries than they do; go will need some chemistry help!
Paula offers to give us some training in how ChEBI curation works - yay! Look into webex session so we can include Harold Drabkin and maybe David Hill. Also consider face-to-face session at (or just before) September GOC meeting.
Chris presents xp example ...
also a brief look at GO MF chebi xps - reactions, most (but not all) from EC; also some binding terms
handle these essentially the same way as BP ones
"ontologizing" ChEBI - make OBO format available
What can GO do now?
- would be good to have a go person dedicated to these xps
- can continue to refer chebi terms, and work on xp representation; leave go-chebi xps in separate file for now
- need to look at all of chebi, so MERIOT not very useful here
- ChEBI will set up request tracker, arrange training, etc.
- GO to make running reasoner part of editing workflow (also see above)
- long term: get funding for someone who can work on GO & ChEBI jointly
- Procedure for GO - use OE reasoner to infer new links; e.g
Full list is in go/scratch/xps/biological_process_xp_chebi-newlinks.txt in cvs
We looked at what has been done so far and revised the editing workflow: Taxon_Editing_Workflow_after_12th_June_2009
Chris is going to set up a cron job to show the errors flagged by the existing links. This runs a script that takes both monophyletic groups and union terms into account. The results are checked into scratch in the go-taxon folder.
Jennifer will use these to inform link addition, and will work through the old sensu terms and the terms flagged by the two GOOSE queries before sending the list or annotations flagged to the annotators for checking.
Lack of has_part is very frustrating, especially in CC. We've had to put in some rather awkward "fake" compositional terms that (a) are painfully verbose and (b) no biologist would ever think of, much less search for. Using has_part would let us disentangle some parts of the graph.
Agreed to test using has_part in a branch file in scratch, and if all goes well, put some in the go-write and go_ext files. (They would be stripped from the downstream files until further notice.)
The has_part relation is defined in RO.
- revive Amelia's plan to write documentation for it
- when documentation is up, announce on GO Friends; include reminder about file options (ext, OBO1.2, OBO1.0; also warn about display considerations)
- use has_part in a branch file in go/scratch/ (e.g. SF 2784316)
- test in OE - can it cope with cycles over has_part + part_of, especially Graph Viewer & Graphviz Viewer
- if OK in OE, roll out into _write and _ext
We'll announce that has_part will be used in CC, then in BP (send reminder just before first BP hp links go in).
Three cases where has_part can be used:
- new terms = new has_part links - no problem!
- existing part_of links that should be removed because they're not really always part_of after all (e.g. U1 snRNA part_of U2-dependent spliceosome; the snRNA is in several splicesome-y complexes) - also no problem, since we'd be correcting errors
- removing part_of links and merging terms to replace with has_part links (e.g. the egregious core TFIIH complex portion of holo TFIIH complex GO:0000439)
- in these cases, check whether the "fake" compositional classes have any annotations; if not, go ahead; if there are annots, contact annot provider ...
- documentation: Amelia
- produce branch file with SF 2784316 edits: Midori
- draft email for gofriends: Midori (circulate among ontdev for edits, then send to gofriends)
- news: Jane
Other new relations
Decided we don't urgently need any new relationship types other that has_part. Spatial relations could enhance CC, but their absence isn't causing trouble.