How to link terms to publications?

[matentzn]

I think there are a few different scenarios to consider on how to link a term to a related publication.

1. A specific annotation assertion, for example, a synonym or definition is sourced from a publication. Here I would suggest we use dc:source. We can agree to use oio:source when using "PMID:25428369"^^xsd:string and dc:source when using https://pubmed.ncbi.nlm.nih.gov/25428369/
2. A more sophisticated discussion on evidence codes that is related can be found here: https://github.com/information-artifact-ontology/ontology-metadata/issues/44

Let us not rekindle the debate about PMID:25428369 vs https://pubmed.ncbi.nlm.nih.gov/25428369/ here. The goal of this ticket is to provide a simple, more meaningful, alternative to this:

HPO:123 oboInOwl:hasDbXref "PMID:123"

[cmungall]

If the intent is to say that the PMID was a source of the term/definition, just annotate the axiom as is standard for most ontologies.

[alanruttenberg]

There's a 'mentions' relation. Not very precise but usable as a start. It's an object property though, not annotations. Although, if there are plans to be able to say more about a paper then having it as an individual. Does ECO have relations? There are a number of older ontology works intended for relating papers, claims and arguments to each other. If that's the direction you are thinking of going in I can look them up.

[matentzn]

Ok, so @srobb1 and the rest of the world:

For now I would recommend:

• If the definition of the term is sourced from some paper, you can follow the convention of the GO lineage like this: ?x IAO:0000115 "My cool definition" [PMID:123], which means, in RDF terms, to put an hasDbXref annotation property on the definition itself (axiom annotation). See GO, MONDO and Uberon for many examples. (for completeness, I don't think this is a conceptually nice pattern, I would prefer dc:source rather than oio:hasDbXref - but it is so widely used, it makes sense to just use it until we reach consensus on how to do it better).

@alanruttenberg @cmungall

mentions relation is arguably not a good relation to curate as part of an ontology - any given term will be mentioned by 1000s of papers so this kind of information is better kept as part of a literature curation in a database. Do you agree?

@srobb1 can you describe what you (and probably others) wanted to express with:

PLANA:123 oboInOwl:hasDbXref "PMID:123"

[alanruttenberg]

We used something like this in the Neurocommons. It's a weak link, but sometimes all you have is that. Re: Databases, a triple store is a database, and the type I prefer to use.

[dosumis]

Why do we need churn on this? It will break downstream use-cases and so at least needs careful co-ordination.

And I'm sorry to be so repetitive - but until someone can tell document how I to reliably use a heterogeneous set of URL to pull metadata on references for an ETL process to KnowledgeBases or other downstream tool - recommending direct use of URLs is basically vandalising processes that already work, however imperfectly.

We can agree to use oio:source when using "PMID:25428369"^^xsd:string and dc:source when using https://pubmed.ncbi.nlm.nih.gov/25428369/