Closed kltm closed 2 years ago
Tagging @pgaudet and @vanaukenk
The choices are:
I'm assuming that 1 or 2 here is what we're currently looking at.
Of course, there is "4.", using a previous release.
@kltm This is very useful. Can you share the error report? Looks like I dont have access to http://skyhook.berkeleybop.org/snapshot
Maybe the ICs are referring to obsolete terms?
Hi - I think this might be an evidence code issue wrt more granular ECO codes like ECO:0005547 mapping up to IC, but the more granular codes don't require a value in the With/From field.
We got an error report of 1506 annotations loaded with NO IC Sample line MGI:104579 Il12rb1 GO:0042022 None Not in the database (translate: MGI id, gene id, GO term used for annotation, IC_id, reason)
The none refers to no value. If it were obsolete , the GO id for the obsolete term would be in the 'none' field, 'Not in the database' would be replaced by 'obsolete' All of the Complex portal annotations we loaded have a blank in our interface.
@vanaukenk I think you are right, ECO:0005547 must be causing the problem.
I didn't realize these were new annotations, Birgit did that before leaving. I think we need to sort this out with ECO. Meanwhile we should load the previous file.
Pascale
For what it's worth, the annotations span a broad range of GO ids for the annotation Here are all of them. GO_InvalidInferredFrom.txt .
I think we thought we had sorted this out with ECO:
https://github.com/evidenceontology/evidenceontology/issues/262
but didn't allow for the consequences of having IC annotations with nothing in the With/From field.
We should discuss what we really want at GO before asking ECO to make more changes.
@pgaudet I did not preserve the report; I'll try and catch it next time around (if we get there). You should have access to skyhook and the reports, but keep in mind that they may not always exist as it gets reset every time there is an attempted run. Assuming that we didn't reset, the report would be available later (my) today.
That said, it looks like the current way forward is to use a previous version ("4", decided https://github.com/geneontology/pipeline/issues/273#issuecomment-1040392170).
The metadata has been updated to the last good upstream source we had for goa_human_complex for a release (see PR above). I'll let it run naturally tonight and try and capture the report for today's soon-to-fail run for reference (if it fits into a gist).
@pgaudet Gist of the goa_human_complex report on snapshot on 2022-02-15 https://gist.github.com/kltm/4df75ce4832e0653219ba7c858582fe0
Thanks @kltm
@vanaukenk diagnosed this correctly - the evidence used by ComplexProtal is considered an IC but is missing 'with'.
There are only about 10 other annotations that fail this rule; I suggest relaxing that rule to a WARNING for now until we figure out what to do about the evidence code. I dont think it's worth stopping the release for this, or excluding all that data.
Thanks, Pascale
This was discusses yesterday on the GO managers' call. @suzialeksander and SGD looked at these errors, and in fact the annotations are experiments done in exogenous systems (more like ISS), but the original source species (what would be in the "with") is not captured. This is against GO rules.
@suzialeksander @vanaukenk Maybe it's OK that these are filtered out?
Thanks, Pascale
@pgaudet @suzialeksander
I would actually prefer we relax the rule to a WARNING and then investigate further.
There is a comment associated with the parent term of the ECO codes used by ComplexPortal that says : "The components in the experimental evidence can come from the same species or a mix of species."
If that is the case for any of the S. cerevisiae complex annotations, then technically they are okay wrt ECO codes.
I'm ok with a warning for now; SGD is discussing if we'll make a hard filter internally or anything today.
Can we close this ticket since the work was done here https://github.com/geneontology/go-site/issues/1794
Work now concentrated on https://github.com/geneontology/go-site/issues/1794; closing
Pipeline failure with
Looks like an over 50% reduction in goa_human_complex:
Looking at the report ( currently http://skyhook.berkeleybop.org/snapshot/reports/goa_human_complex-report.html#gorule-0000016, but will be reset when
snapshot
tries again tonight), there were 3151 violations of GO Rule 16 (IC with/from reqs). E.g.