Closed hdrabkin closed 3 years ago
Hi @hdrabkin thanks for reporting the issue, I just checked and its loading the model as below. However there are few problems to address to @ukemi @vanaukenk @lpalbou . It seems like all the GPs in this particular model do not have root types, i.e not information biomacromolecule nor any CHEBI. Maybe they are not loaded in NEO. Also as you can see, some MGI ids are not resolved into their respective labels.
However @hdrabkin I don't think PMID errors is the issue? So far I think it displays PMIDs with wrong format (space) but only not searchable. The issue is that since some GPs are not recognized as GPs for above reasons, the whole evidence on the enabled by edge is not read thus empty evidence. Once the NEO issue is fixed everything will show properly
@vanaukenk It seems like we should move this to the NEO tracker?
I mentioned the PMID because THAT is what Dmitry tried to fix because it's not the correct PMID; the annotations have loaded into MGI; we wanted to fix the PMID using ART but could not.
@kltm @tmushayahama @balhoff
I'm not sure I understand what's going on with this particular model. There looks to be annotation to protein-coding genes but it doesn't appear that MGI models containing annotations to protein-coding genes universally have this problem.
@tmushayahama Can you confirm, for example, that the entities in gomodel:60e7989d00000247 do have associated root types?
@ukemi @hdrabkin __
The model also looks a bit weird. It looks like some of the connections between generic MFs and BPs didn't save?
Do you want to use this for trouble-shooting or do you want me to go in and fix the refs? They are easy to find. What's the right PMID?
But now I can't even save the model. Seems like we have run into this before.
Ugh - is this the model we've had trouble saving in the past?
It might be. Looking at it, it almost looks like at some point it partially saved.
Hi - I just checked my notes from the various Noctua outages, and indeed, this is the model we've had trouble with in the past.
I know there are a lot of annotations there, but we may have to cut our losses and create a new model and mark this for deletion.
So this is really curious then because I looked at this model before and I didn't see all those 'anonymous' nodes. If we create a new one, we should export the GPAD so we have a record of what's there and the annotations can be re-entered. Should we encourage splitting giant 'single annotation' models like this into smaller chunks?
We could at least change it to dev, so that it won't be output to our gpad; During the drop/reload, there will be no annotations using the offensive PMID to that it would now not have any data in MGI and could be deleted.
Dmitry will have the right PMID
I think we should have him export the GPAD from this model and then one of us should try to delete the model. Then we can use the GPAD to re-enter the data into a new model.
There are a couple strategies we could use. Let's discuss at lab meeting tomorrow.
Seems like there are several issues here.
GP not showing in ART. Most of them look valid and should show up in ART. Possible issues: 1) ART; 2) Discrepancies between the GOLr used to select a GP in Noctua graph and the one used to resolve a GP in ART; 3) if partially saved, this might have impacted the GP identifier and prevent the resolving of the root type / name of the gene
Why this model is not saving. @kltm we would need the log for that to try to track down what's happening, if it's a memory issue, if it's a request just too big and crash the network or timeout, etc
@vanaukenk to reply https://github.com/geneontology/neo/issues/73#issuecomment-880116623 , no they don't have root types
Tagging @balhoff on this one 'Traf3ip2' seems to autocomplete but not appearing in root types. Is there something needs to be done on Minerva?
This model still has that weird backquote in it. I would export the OWL from the UI, delete that character in a text editor, manually commit this to github, and then reload models in Minerva. I don't think you need to lose anything.
I've exported the GPAD form this model and saved it on my machine. No matter what we do, the annotations will not be lost. There are annotations to this gene:
MGI MGI:2143599 involved_in GO:0042493 PMID:19155511 ECO:0000315 MGI:MGI:3831153 20210121 MGI noctua-model-id=gomodel:5fce9b7300002524|contributor=http://orcid.org/0000-0003-3394-9805|model-state=production MGI MGI:98283 involved_in GO:0048255 PMID:21822258 ECO:0000314 20210128 MGI has_participant(MGI:MGI:101835),has_participant(MGI:MGI:107548),has_participant(MGI:MGI:2143599),has_participant(MGI:MGI:98283) noctua-model-id=gomodel:5fce9b7300002524|contributor=http://orcid.org/0000-0003-3394-9805|model-state=production MGI MGI:2143599 involved_in GO:0042092 PMID:21856933 ECO:0000315 MGI:MGI:3510784 20210217 MGI part_of(GO:0006954),has_participant(MGI:MGI:2155888),has_participant(CL:0000084),has_participant(CL:0000546) noctua-model-id=gomodel:5fce9b7300002524|contributor=http://orcid.org/0000-0003-3394-9805|model-state=production
Okay, this seems like a dupe of https://github.com/geneontology/minerva/issues/379
@kltm should you and I coordinate on a Minerva reload and restart time to take care of this?
@balhoff I'm making notes on the other ticket
that weird backquote in it
Question from left field - is this the sort of problem that could arise if a curator tried to copy-paste text straight from a Word document with all its wonderful non-ASCII quote characters, into a Noctua form? So a long-term fix is still more curator training, emphasizing the need for plain text?
@deustp01 Could be, although in this case, it may be from the more permissive graph editor Long term, fixes will be a combination of:
https://github.com/geneontology/minerva/issues/379 was originally formulated with a bit of the last one in mind.
that weird backquote in it
Question from left field - is this the sort of problem that could arise if a curator tried to copy-paste text straight from a Word document with all its wonderful non-ASCII quote characters, into a Noctua form? So a long-term fix is still more curator training, emphasizing the need for plain text?
I don't think it's possible because the value entered in Noctua Form is the resolved ids not the copied text as we are using autocomplete. I think it is an upstream problem. Lemme double check
gomodel:5fce9b7300002524
We know there are several errors with incorrect PMID (PMID: 28581022 ) used for annotation to Il17a, Il17ra, Notch1, and Traf3ip2 that need to be corrected, which would be easier in ART, but it will not load. We can see gpad output from graphic view, but cannot get annotation preview The annotations are coming into the MGI gpad and load into MGI
Is there something specific that is preventing loading into ART?