Include Uberon 'part of' relations (possibly for disease relationship inference)

[lschriml]

Allen: Sue noticed that endocrine gland is no longer a child of endocrine system. Had a look at UBERON in OLS, and the relationship is that file. Had a look in diid-edit.owl file and we do not have 'endocrine gland' as a child of 'endocrine system' Also noticed that the uberon system terms do not have children. Could this be from a step that removed their EQ/SubClass relationships OR are we pulling a file that is not rendered (where all inferred relationships are seen as asserted) ?

Cheers, Lynn

[sbello]

@lschriml I'm not sure if this is new or if it just popped up now due to new annotations. I think the issue may be because endocrine gland has only a part_of relationship to endocrine system and the part_of relationships don't appear to be in the import file.

[allenbaron]

The relationship between 'endocrine gland' and 'endocrine system' is missing because it is not asserted and only the asserted tree is maintained in the import. It's also further complicated because of the the 'part of' axiom it's using.

It's not a new issue. The relationship between endocrine gland and endocrine system is also missing from the uberon_import.owl (2022-03-02 release), only the asserted subClassOf relationship to UBERON_0002530 ("gland") is present. https://github.com/DiseaseOntology/HumanDiseaseOntology/blob/d57c42e99e9ae27783c701edf1ca12569bcc9def/src/ontology/imports/uberon_import.owl#L3877-L3883

EMBL's OLS shows both subClassOf and 'part of' relationships (see image, 'P' indicates 'part of').

Can disease-ontology.org or our downstream users (like MGI) use or display 'part of' relationships?

If we want to have these in the uberon_import.owl file, we could update our import build procedure to keep these statements (I think). But if there's not a way for us or our major users to visualize/use 'part of' relationships, we'd have to go a step further and either add this functionality to our loaders/systems OR write something custom to convert the 'part of' relationship axioms to non-axiomatized subClassOf statements. I'd be very hesitant to do the latter because it means re-writing Uberon on a significant scale.

[sbello]

Hmm, the MP imports built using the ODK include the "part of" relations DO may want to adjust the imports to include part of's at least for anatomy where the part of relationship is often the most informative

[allenbaron]

I tested importing the 'part of' relationship and it had no effect on reasoning. On OBO slack Jim Balhoff said it would probably require a property chain axiom such as 'disease has location' o 'part of' -> 'disease has location', which he thought was reasonable.

Should we propose this change in RO?

[allenbaron]

Concrete example for later testing: Inferring ‘endocrine gland cancer’ (DOID:170) as a subclass of ‘endocrine system disease’ (DOID:28)

Details:

• ‘endocrine gland cancer’ (DOID:170) has the EqClass axiom: cancer and ('disease has location' some 'endocrine gland')
• ‘endocrine system disease’ (DOID:28) has the EqClass axiom: disease and ('disease has location' some 'endocrine system')
• cancer (DOID:162) is a subclass of ‘disease of cellular proliferation’ (DOID:14566) which is a subclass of ‘disease’ (DOID:4)
• In Uberon, ‘endocrine gland’ (UBERON:0002368) has the subclass axiom: 'part of' some 'endocrine system'

[lschriml]

@allenbaron -- Can this ticket be closed ?

[allenbaron]

We have not completed it yet.