MGI xrefs failing GO checks

[kltm]

From @ValWood at pombase/pombase-chado#1224

Our MGI ISO xrefs are failing checks.

WARNING - Invalid identifier:GORULE:0000027: 1298204 does not match any id_syntax patterns for MGI in dbxrefs--PomBase SPBC530.12c pdf1 enables GO:0008474 PMID:15075260 ISO MGI:MGI:1298204 F palmitoyl protein thioesterase/ dolichol pyrophosphate phosphatase fusion protein Pdf1 protein taxon:4896 20040414 PomBase 
WARNING - Invalid identifier:GORULE:0000027: 1316717 does not match any id_syntax patterns for MGI in dbxrefs--PomBase SPBC20F10.03 SPBC20F10.03 is_active_in GO:0005634 GO_REF:0000024 ISS MGI:MGI:1316717 C armadillo-type fold protein, human IFRD1 ortholog, implicated in transcription or signaling protein taxon:4896 20170830 PomBase 
WARNING - Invalid identifier:GORULE:0000027: 1346084 does not match any id_syntax patterns for MGI in dbxrefs--PomBase SPAC6C3.09 rpp40 part_of GO:0005655 GO_REF:0000024 ISS MGI:MGI:1346084 C RNase P and RNase MRP subunit Rpp40 protein taxon:4896 20061017 PomBase 
WARNING - Invalid identifier:GORULE:0000027: 1919005 does not match any id_syntax patterns for MGI in dbxrefs--PomBase SPAC513.06c dhd1 involved_in GO:0042843 GO_REF:0000024 ISS MGI:MGI:1919005 P D-xylose 1-dehydrogenase (NADP+) protein taxon:4896 20150502 PomBase 
WARNING - Invalid identifier:GORULE:0000027: 1919005 does not match any id_syntax patterns for MGI in dbxrefs--PomBase SPAC513.06c dhd1 enables GO:0047837 GO_REF:0000024 ISS MGI:MGI:1919005 F D-xylose 1-dehydrogenase (NADP+) protein taxon:4896 20150502 PomBase 

but its a bit weird because the display and the URL are MGI:1298204 but on the pop-up it says MGI:1298204
could you have a dig and see if the syntax has been resolved to remove the first MGI: or something?

@kltm 's response:

Looking at https://github.com/geneontology/go-site/blob/master/metadata/rules/gorule-0000027.md . Okay, "soft" warning, so no data filtering.

The moment of failure is likely here:
https://github.com/biolink/ontobio/blob/master/ontobio/io/assocparser.py#L835
Special casing for MGI leading into it is:
https://github.com/biolink/ontobio/blob/master/ontobio/io/assocparser.py#L802-L806

So, it looks like MGI:MGI:1919005 would be clipped to MGI and 1919005, the latter of which would fail when checking against the regexp. The options here would be:

• change the dbxrefs regexp to reflect our behind-the-scenes fix of MGI (I'm not sure what the knock-on effect would be)
• remove the ontobio "fix" (I'm not sure what the knock-on effect would be)
• change the MGI full id to MGI:MGI:MGI:1919005 (I know what the knock-on effect would be: hilarity)

Either way, @pgaudet , this is probably best approached as a GO QC bug for the moment (although a "light" one as no fix or filtering is done) and added to the QC worklist.

[kltm]

@pgaudet I've temporarily put this in the "low-hanging fruit" project in the spec and prioritize section.

[mugitty]

From db.xrefs file:

- database: MGI
  name: Mouse Genome Informatics
  rdf_uri_prefix: http://identifiers.org/MGI/
  generic_urls:
    - http://www.informatics.jax.org/
  entity_types:
    - type_name: gene
      type_id: SO:0000704
      id_syntax: MGI:[0-9]{5,}
      url_syntax: http://www.informatics.jax.org/accession/[example_id]
      example_id: MGI:MGI:1345277
      example_url: http://www.informatics.jax.org/accession/MGI:1345277
    - type_name: variation
      type_id: VariO:0001
      id_syntax: MGI:[0-9]{5,}
      url_syntax: http://www.informatics.jax.org/accession/[example_id]
      example_id: MGI:MGI:3590672
      example_url: http://www.informatics.jax.org/accession/MGI:3590672

The code validates against the id_syntax field. A new entry has to be added with
id_syntax: MGI:MGI:[0-9]{5,}

[kltm]

@mugitty Hm. I think that entry is actually correct. If you look at something like SGD or WB, you can see that they match in intention, even though MGI is a special case. (The field names id_syntax and example_id are very poor; they should be: internal_id_syntax and example_curie or something...)

[mugitty]

@kltm, the code is using id_syntax, which is specified in https://github.com/geneontology/go-site/blob/master/metadata/db-xrefs.schema.yaml. If internal_id_syntax is to be used then it has to be added to db-xrefs.schema.yaml. Currently, id_syntax does not match example_id. The code handles multiple id_syntax entries. It does not use example_id

[kltm]

@mugitty Yes, that is correct: id_syntax never matches example_id, as they are different concepts.
Take a look at another example, like SGD: https://github.com/geneontology/go-site/blob/master/metadata/db-xrefs.yaml#L2386-L2399 . id_syntax never matches example_id, as they are kinda misnamed for historical reasons.

id_syntax: the regexp for the database's internal id
example_id: an example CURIE (namespace plus internal id) of the resource

If the metadata is correct, we need to look at affecting the change we want--with the metadata we have--in the code. The MGI:MGI doubling has always caused problems...

[mugitty]

@pgaudet, the internal representation of MGI was updated due to geneontology/go-site#91

I will update to handle what is in db-xrefs.yaml as well as the internal representation