Issues with abstract terms: interacts_with

[colleenXu]

@sierra-moxon

This issue is to focus on "interacts with". We in Service Provider want guidance: should "interacts with" be an abstract term? Are there better biolink 3.0.3 terms for these cases? (3.0.3 is what we plan to migrate to, but knowing about future version terms is also helpful)

Service Provider currently uses "interacts with" in the following ways:

1. used for Monarch's gene-gene interaction info. It's not clear if biolink "genetically interacts with" matches because its definition seems too specific.
   • For example, the DNAH5 - PLEC association links to this publication and biogrid as a source. It's not clear what the relationship is (protein-protein interaction? possible epigenetic/histone-related relationship?).
   • Also, biogrid includes all sorts of interactions with all kinds of certainty levels.
2. used for DGIdb's drug-gene interaction info that is categorized as "other/unknown" or "n/a" - this is ~76% of the data we have from this resource.
3. used for iDISK's supplement-drug interaction info
4. used for SEMMEDDB's data with the predicate SEMMEDDB:INTERACTS_WITH. This semmed term is mapped to biolink "interacts with", so this mapping should be changed if users of biolink-model are not supposed to annotate data with "interacts with"

Note: for our dev instance (already on biolink 3.0.3), we have already changed cases 1 and 2 to use "related to".