Links to TAIR locus pages are wrong in AmiGO

[kltm]

From @tberardini on May 23, 2018 15:10

Example:

http://amigo.geneontology.org/amigo/gene_product/TAIR:locus:2007651

The link URL to TAIR generated by AmiGO for locus:2007651 is

http://arabidopsis.org/servlets/TairObject?type=communication&id=locus:2007651

But should be:

https://www.arabidopsis.org/servlets/TairObject?accession=locus:2007651

Please fix. Thank you.

Copied from original issue: geneontology/amigo#503

[kltm]

This likely traces back to #443

[kltm]

@cmungall This seems similar to what we went through with WB a couple of months ago. This is essentially a collision being expressed in the "linker" code as it uses the main namespace as the lookup and TAIR has two different possible target for the same namespace.
See https://github.com/geneontology/go-site/blob/master/metadata/db-xrefs.yaml#L2892
As, I believe, is mentioned in the analogus WB tickets from April, full regexp lookup for all links would be quite expensive and we went with suggesting either splitting the namespace and/or having the receiving resource deal with the local id part. Experimenting with TAIR, the latter does not seem to be in place.

[kltm]

Also note: #158

[tberardini]

Anything I can do to help resolve this?

[cmungall]

sorry can't give full attention now, can we just have the gene link be the default for now? This is 99.9% of the links clicked on I would wager

[tberardini]

can we just have the gene link be the default for now?

Yes please.

[kltm]

Will do.

[kltm]

This "patch"/"hack" is now in place for TAIR.

[tberardini]

Thank you!

[kltm]

NP.

I just want to spell this out so when we pick up on this again we have a quick start.
So, for example, here is all of the TAIR-ish information in our metadata as of this moment:

- database: AGI_LocusCode
  name: Arabidopsis Genome Initiative
  description: Comprises TAIR, TIGR and MIPS
  generic_urls:
    - http://www.arabidopsis.org
  entity_types:
    - type_name: gene
      type_id: SO:0000704
      id_syntax: A[Tt][MmCc0-5][Gg][0-9]{5}(\.[0-9]{1})?
      url_syntax: http://arabidopsis.org/servlets/TairObject?type=locus&name=[example_id]
      example_id: AGI_LocusCode:At2g17950
      example_url: http://arabidopsis.org/servlets/TairObject?type=locus&name=At2g17950
- database: AraCyc
  name: AraCyc metabolic pathway database for Arabidopsis thaliana
  generic_urls:
    - http://www.arabidopsis.org/biocyc/index.jsp
  entity_types:
    - type_name: entity
      type_id: BET:0000000
      url_syntax: http://www.arabidopsis.org:1555/ARA/NEW-IMAGE?type=NIL&object=[example_id]
      example_id: AraCyc:PWYQT-62
      example_url: http://www.arabidopsis.org:1555/ARA/NEW-IMAGE?type=NIL&object=PWYQT-62
- database: TAIR
  name: The Arabidopsis Information Resource
  rdf_uri_prefix: http://identifiers.org/tair.locus/
  generic_urls:
    - http://www.arabidopsis.org/
  entity_types:
    - type_name: gene
      type_id: SO:0000704
      id_syntax: gene:[0-9]{7,12}
      url_syntax: http://arabidopsis.org/servlets/TairObject?accession=[example_id]
      example_id: TAIR:gene:2062713
      example_url: http://arabidopsis.org/servlets/TairObject?accession=gene:2062713
    - type_name: communication
      type_id: BET:0000000
      id_syntax: Communication:[0-9]{7,12}
      url_syntax: http://arabidopsis.org/servlets/TairObject?type=communication&id=[example_id]
      example_id: TAIR:Communication:1345790
      example_url: http://arabidopsis.org/servlets/TairObject?type=communication&id=1345790
    - type_name: primary transcript
      type_id: SO:0000185
      id_syntax: locus:[0-9]{7}
      url_syntax: http://arabidopsis.org/servlets/TairObject?accession=[example_id]
      example_id: TAIR:locus:2146653
      example_url: http://arabidopsis.org/servlets/TairObject?accession=locus:2146653

That's four URL schemes under three namespaces. Ideally, what we'd want is one of two things, which practically boils down to one thing done two different ways: a one namespace per URL scheme.

1. Split the TAIR communications into their own namespace (e.g. TAIR_COMM), like PAINT_REF, etc.
2. Setup a TAIR endpoint to resolve local IDs such that all entity types here have the same URL scheme (e.g. http://arabidopsis.org/servlets/TairObject?accession=ANYTHING:LOCAL_WHATEVS).

Really, for what we're doing, a better metadata format would have be to have the url_syntax stuff be at a higher level than under entity_types so that there was no confusing repetition for something we expect to be the same in practice.

[tberardini]

Came here from a TAIR helpdesk email. After reviewing the SO mapping for the entity_types above, I think we need to edit a bit. In TAIR, locus = SO:gene and gene = SO:primary_transcript.

- type_name: gene

SHOULD BE primary_transcript

  type_id: SO:0000704

SHOULD BE SO:0000185

  id_syntax: gene:[0-9]{7,12}
  url_syntax: http://arabidopsis.org/servlets/TairObject?accession=[example_id]
  example_id: TAIR:gene:2062713
  example_url: http://arabidopsis.org/servlets/TairObject?accession=gene:2062713

- type_name: primary transcript

SHOULD BE gene

  type_id: SO:0000185

SHOULD BE SO:0000704

  id_syntax: locus:[0-9]{7}
  url_syntax: http://arabidopsis.org/servlets/TairObject?accession=[example_id]
  example_id: TAIR:locus:2146653
  example_url: http://arabidopsis.org/servlets/TairObject?accession=locus:2146653

[dustine32]

@tberardini Thanks for the response! Yep, this was the first approach that @thomaspd and I were thinking but we're a little hesitant to claim all TAIR genes follow this. Just letting you know making this change would likely invalidate any TAIR entities in an annotation's "with/from" field if the identifier followed the "gene:1234567" syntax. Would this be OK?

[tberardini]

Just letting you know making this change would likely invalidate any TAIR entities in an annotation's "with/from" field if the identifier followed the "gene:1234567" syntax. Would this be OK?

Is that because the with/from field requires the entity to be in a subset of classes that don't include primary_transcript?

[dustine32]

@tberardini Correct! So just throw primary_transcript in this subset as well?

[tberardini]

@dustine32 I would say, yes.

[pgaudet]

@tberardini @dustine32 Is this fixed ?

[tberardini]

I've given my input, hope that was sufficient.

[kltm]

@pgaudet As far as I know, what is outlined in #659 (comment) has not been completed (as far as what closure would be). It does not seem to be pressing at this point?

[pgaudet]

Action point: Need identifier best practices when interacting with GO.

[cmungall]

I am closing this because the original issue is resolved

however, I have many questions here and will open a new issue