search

information-artifact-ontology / ontology-metadata

OBO Metadata Ontology
Creative Commons Zero v1.0 Universal
19 stars 8 forks source link

Specify the mapped/equivalent terms #31

Open zhengj2007 opened 6 years ago

zhengj2007 commented 6 years ago

Most commonly used approach is using annotationProperty: database_cross_reference (example: http://purl.obolibrary.org/obo/DOID_9351). When multiple mapped terms need to be listed, what is the correct format we should use to separate these mapped terms? Is this approach applied to both OBO Foundry ontologies and other ontologies/terminologies, like SNOMED, ICD-9?

Some used class equivalent, like GO:cell = CL:cell, when should we take this approach? Does it only apply to OBO Foundry ontologies?

Some used rdfs:seeAlso annotationProperty, see example: http://purl.obolibrary.org/obo/VO_0000609, (Note: it may not the correct annotationProperty to use.)

Some used self-defined annotationProperty, see example: http://purl.obolibrary.org/obo/DRON_00016163

Related issue: https://github.com/information-artifact-ontology/ontology-metadata/issues/24

Comments from Chris M: Loosely I think we should encourage in order:

We should discourage db-specific prefixes

cmungall commented 6 years ago

xref has no semantics. It's not that it isn't useful, it's just weaker than equivalence. If you know it's equivalent, use an equivalence axiom. If you don't know then don't. In some cases we shadow an equivalence axiom with an xref, for applications that want xrefs. See notes for obo version of obofoundry.org/ontology/mondo.html. Also I believe OxO expects xrefs, cc @simonjupp

I would say seeAlso is even weaker. It is also typically used with a URL not a CURIE. E.g. it may link a class to a page about that class.

zhengj2007 commented 6 years ago

Ramona: In BCO, we use OWL same as, and use an imported ontology to specify those axioms so they can be included or not as needed.

zhengj2007 commented 6 years ago

Discussed on Mar 27, 2018 call. @cmungall will make some examples and how they deal with them now.

jonquet commented 6 years ago

Glad to see the discussion here. Within the AgroPortal project with @elcioabrahao we ware currently working on building a mapping repository and therefore extracting mappings from the ontology source files. Honestly working with XREFS is a bit of a nightmare ;)

We will share a document soon that groups the results per ontology and show the number of unresolvable targets. XREFS are supposed to be manually created... thus they are supposed to be good and serve as reference for other (automatic) ontology alignement solutions ;)

I agree with @cmungall point before and included in original post. If we could encourage a bit more the community to adopt standard mapping properties that would be very good. owl:sameAs, SKOS ones (https://www.w3.org/TR/2009/REC-skos-reference-20090818/#mapping) and finaly, rdfs:seeAlso (that are the weakest indeed).

As a recall the OBO spec provide headers for XREFS: http://owlcollab.github.io/oboformat/doc/obo-syntax.html (section 4.4.2) We have found some but not sure how to use them yet. I will open another tracker to discuss possible equivalence with standard mapping properties.

So how can we do to improve /curate the XREFS?

cmungall commented 6 years ago

See also: https://github.com/ontodev/robot/issues/312

https://github.com/dcppc/Metadata_matrix/issues/2

matentzn commented 4 years ago

So while have a bit of a plan to alleviate the xref madness in the future, I am not sure why the database_cross_reference property is not part of OMO.. Is this by design?

jonquet commented 4 years ago

Hi all, we have an ICBO submission under review on our work on XRefs in the OBO world which will be highly relevant for this thread. I can send it offline for the moment. We come up with an analysis and some recommendations.

cmungall commented 3 years ago

What is the update here?

It is clear that omo needs to include both oio:hasDbXref and the 4 skos match predicates

matentzn commented 3 years ago

Agree with that @cmungall