Check for existence of conflicting category assignments

[caufieldjh]

A graph may be assigned multiple Biolink (or other) categories, but these should adhere to a model.
Including both a child and a parent category is not a violation of the model.
Including two or more categories exclusive of each other (like biolink:Disease|biolink:Device) would be a violation.
But how do we define this?
At minimum, the category should be the same as that of the type entity.
If a type entity isn't available, we should be able to find a path between the two categories without involving a root.

[matentzn]

I would be curious to understand what the biolink vision for this is.

[cmungall]

Note that some people in Translator actively want conflation between e.g. a gene, the corresponding reference protein, and the corresponding OMIM mendelian disease entry, which are often all in 1:1. I don't think this is good, but biolink is currently quite lenient when it comes to formal disjointness declarations.

For our purposes, we want to treat the main is-a hierarchy as pairwise disjoint.

of course, something can be both a pathway and a mixin like OntologyClass

even in cases where we do want to conflate e.g gene to refP, we should pick one, so we should treat the is-a hierarchy as pairwise disjoint