Ontology meeting 2014-10-09
Attendees: Jane, Paola, DavidH, Chris, Heiko, Harold, Tanya, Judy
TG template for 'cell migration'
What is the most appropriate relation to use? See email thread '[go-ontology] cell migration templates'
Mix of 'has_input,' 'results in movement of,' and 'RO:0002313 transportsormaintainslocalizationof'. Action items: Check what UniProtKB has used in annotation extension, MGI uses 'results in movement of', update ontology to replace 'has_input' with 'results in movement of' in the 'cell migration' children terms. Heiko to change relationship type in TermGenie for this template.
GO into NCIt
No next steps yet. NCIt will experiment with a small subset of GO and doing the swap. Google Doc has info. NCIt has pathological processes. These will not be in GO. More discussion necessary, hopefully before the Barcelona. NCIt is in UMLS Metathesaurus, there should already be cross-references, there may be incomplete. Project still in speculative state.
Follow-up on negatively_regulates o part_of/positively_regulates o part_of
We'll go ahead with these, right? Previous discussion copied below:
(Need them for SF 11166)
Chris made a correction. The new diff is attached to his email with subject 'reasoner diff before and after' and dated Sept. 25th. The list is quite long - did anyone get a chance to review it?
We also need to think of this in the context of the call on monday. See: http://go-genkisugi.rhcloud.com/seed/model/gomodel:542732ad0000005 Would NOS inhibitor activity classify under the negative regulation term here? Because that is implied by the classic GAF annotation
AI: Let's give it another week and look at at least sections of it... we will all look at random sections of it and comment/decide at next week's call. If we haven't had time, we'll go ahead all the same.
Discussion on direct/indirect regulation. AI: Some of us will look at examples and figure out suggestions. David H will submit Nikolai's work as an RO ticket, and will check his presentation/slides into svn.
Problem: Redundant is_a/part_of links removed by DavidH. Part_of were removed. Is_a's should remain. Future rule to follow: All the parts of a regulator process are actually regulatory processes, so the relationship should be is_a and not part_of. Action: Inform people about these new inferred links (over 1000) using the GO website.
Question: 'Redundant' inferred A pos/neg regulates B and asserted A regulates B. Remove the less specific and keep the more specific. There will be a lot of these. Not a huge priority but noted because it can confuse things and clutter up the graph.
Question: What to do with 'in/directly activates' and 'in/directly inhibits' and property chains? Use Nikolai's approach?
Automatic Annotation Deepening
Also know as Fold-based prediction. This is currently the only 'experimental' prediction method.
Idea: use c16 information and equivalent class axioms to check for more specific classes. For example
GO:0008283 cell proliferation + acts_on_population_of(CL:0000037 hematopoietic stem cell) ==> GO:0071425 hematopoietic stem cell proliferation
GO:0009913 epidermal cell differentiation + results_in_acquisition_of_features_of(CL:0002187 basal cell of epidermis) ==> GO:0030216 keratinocyte differentiation
GO:0030182 neuron differentiation + results_in_acquisition_of_features_of(CL:0000617 GABAergic neuron) ==> GO:0097154 GABAergic neuron differentiation
GO:0001101 response to acid chemical + has_direct_input(CHEBI:3098 bile acid) ==> GO:1903412 response to bile acid
What about uninstantiated classes? We would only deepen to the most specific term that exists in GO. Could theoretically have some kind of TG-like procedure. May be too complicated for now = Analysis Ontology with fake GO id with right placement in graph. Kind of exists. Use an intermediate file in Term Enrichment to make it as good as it can be.
- goa_human http://build.berkeleybop.org/view/GAF/job/gaf-check-goa_human/
- MGI http://build.berkeleybop.org/view/GAF/job/gaf-check-mgi/
- Pombase http://build.berkeleybop.org/view/GAF/job/gaf-check-pombase/
- SGD (soon)
Look at 'gaf-prediction-experimental-report.txt' and 'gene_association.mgi.inf.experimental.gaf'
See SF issue 11241 In general, if a chemical is used as an analog of the biologically relevant chemical for an assay of MF/BP, should the corresponding MF/BP be represented by GO terms?
Editor's discussion: Removal of 'assay' terms as we discover them. Annotate to the intended transported molecule, in this case, K+. Then add the OBI, BAO, or bioassay term - measure of K+ transport using rubidium, in the with field. (IDA + evidence with allowed where with field contains the assay term).
TG new curators?
Today, got a request to be a submitter in TG from Penelope Garmiri (UniProt Knowledgebase Content Team) What is the SOP?
If Heiko receives a request by an 'unknown' person for TG permissions, this person needs to be approved by someone on GO Eds before granting access.