NTR: "dbxref prefix map"

[matentzn]

I am trying to make a dent into the non-interpretability of the prefixes used in hasDbXref expressions. As discussed in great detail elsewhere, hasDbXref will continue to be a CURIE string (not an entity) that lives outside of the normal RDF world. In order to standardise uses of prefixes, I want to propose at least an AP that we can use to link to a context we can use to disambiguate the prefixes used in the hasDbXref properties of some ontology. So I envision an annotation like:

obo:my.owl rdf:type owl:Ontology .
obo:my.owl OMO:0001111 <https://raw.githubusercontent.com/geneontology/go-site/master/metadata/db-xrefs.yaml> .

This could then tell the user that

my:001 oio:hasDbXref "HP:001"

refers to and HPO term.

Or better yet, linking it to a formal context.

Anyone opposed to such a property?

[matentzn]

cc @dosumis @cmungall

[dosumis]

Agreed.

1. We should tag anything with a valid OBO prefix as an ontology mapping and not subject this to QC.
2. For all the others, we should look up in GO_xrefs yaml and pull the appropriate stanza from there if possible. Remaining should be curated or deleted.

Also agreed that the master mapping file can lives on the OMO repo.

[balhoff]

An alternative approach might be to include the prefix declarations in machine-readable form (but "reified") within the ontology rather than linking out, using the SHACL vocabulary. We're using this here: https://github.com/OBOFoundry/OBOFoundry.github.io/blob/master/registry/obo_prefixes.ttl

[matentzn]

@balhoff totally, good idea. Would you say, just importing this with owl:imports?

[balhoff]

@balhoff totally, good idea. Would you say, just importing this with owl:imports?

Yeah that sounds great.

[matentzn]

Here is the current bioregistry shacl prefixmap: https://github.com/biopragmatics/bioregistry/blob/main/exports/contexts/obo_synonyms.context.ttl

This brings us only half-way. What remains is how we can connect this information with other validation information, such as ID regex patterns?

[cthoyt]

Does shacl have a way of representing the regex? Or can we introduce a new property to go along with shacl where we can add regexes into this bioregistry export?

Update: indeed it does! See https://github.com/biopragmatics/bioregistry/blob/main/exports/contexts/bioregistry.context.ttl for an example of the Bioregistry output in shacl that includes sh:pattern for many resources.

[cthoyt]

Update, SHACL does have a way to represent patterns. See https://github.com/biopragmatics/bioregistry/blob/main/exports/contexts/bioregistry.context.ttl for an example.