search

monarch-initiative / mondo

Mondo Disease Ontology
http://obofoundry.org/ontology/mondo
Creative Commons Attribution 4.0 International
229 stars 52 forks source link

xrefs vs basicPropertyValues exactMatch #5430

Open MichaelaEBI opened 2 years ago

MichaelaEBI commented 2 years ago

What is the difference between xrefs for a Mondo disease term and the basicPropertyValues field with "pred" : "http://www.w3.org/2004/02/skos/core#exactMatch"? I have looked in the FAQs and documentation, but did not find this information. For context, I was looking at MeSH identifiers in these two fields and for 8,036 Mondo terms the MeSH entry seems to appear in both places, but for 42 Mondo terms the MeSH ID only appears as xrefs (I used the csv files from data.bioontology.org). Here is one example showing the 'Preferred Label', 'xrefs' and 'exactMatch' entries:

benign duodenal neoplasm
DOID:1737|EFO:1000907|MedDRA:10004251|SCTID:92080005|MESH:D004379|NCIT:C4775
http://identifiers.org/snomedct/92080005|http://purl.obolibrary.org/obo/DOID_1737|http://purl.obolibrary.org/obo/NCIT_C4775

Further, how is the 'Other mappings' field on the website populated?

nicolevasilevsky commented 2 years ago

Hi @MichaelaEBI looks like the othr mappings in Monarch are the equivalent database cross references in Mondo (@putmantime can you confirm this is how Monarch is displaying these dbxrefs?)

We do have this FAQ on the website abut exact match, which I pasted below, but it sounds like this does not answer your question?

What is the difference between MONDO:equivalentTo and skos:exactMatch?

MONDO:equivalentTo and skos:exactMatch conceptually overlap but they have entirely different semantics. Mondo:equivalentTo is much stronger than skos:exactMatch. Mondo uses skos:exactMatch to bridge semantic spaces which conceptually overlap, but are not (necessarily) logically coherent according to OWL semantics. For example, Mondo defines X (an example here of a disease) which is an exact match to Y (add mapped MeSH term) in MeSH, which is not an OWL ontology - trying to apply an equivalent relation with OWL semantics would not make much sense.

Moreover, merging Mondo with another disease ontology, even if both are maintained in OWL, is not guaranteed to produce a coherent result, i.e one that contains no logical errors. This is a very fundamental feature of Mondo: integrating various disease classification to create a harmonised classification. This harmonised classification may, by design, partially disagree with any particular source. Therefore, applying strong OWL semantics with owl:equivalentClass is, usually, inappropriate.

@matentzn and @cmungall are either of you better able to answer the question above?

@MichaelaEBI our team may need to discuss this further on a call and it may take us some time to get to this agenda item. I'll try to address this asap. :)

matentzn commented 1 year ago

@nicolevasilevsky I think this documentation entry needs to be reworked a bit.

What this question suggests to me is that we should be more deliberate with what we export during a release. Right now MONDO:equivalentTo are all translated into skos:exactMatch (at least for most of the source terminologies).

When this documentation you shared was active, we used to export owl equivalent class axioms (which we still do in "mondo-with-equivalents.owl" release), but nowadays, this does not make sense anymore - we do not at all suggest the use of owl semantics anymore, just the more "loose" interpretation of skos.

However, @MichaelaEBI the answer, if you rewrite slightly, still applies:

skos:exactMatch has a much stronger semantic meaning than xref. From a user perspective I would treat entries with xref as "rough matches", while the skos:exactMatch ones as precise matches.

sabrinatoro commented 4 months ago

@nicolevasilevsky is there anything else to do to address this issue?

matentzn commented 3 months ago