RO terms that don't normalize

[cbizon]

The Edge Normalization service tries to find a biolink model term that can be used for the possibly more specific curie input. The first thing we do is check slot_uri and mappings, to see if we find any exact matches. Failing that, if the input is an RO terms, we start moving up to RO superproperties looking for a match in biolink.

However, not everything in RO comes together into a very generic "related_to" apex (while in biolink model it does). So that means we get up to the top of an RO branch and still never find a biolink model match. For edge normalization, we just return "related to" and call it a day, but the question remains: should any of these terms have direct mappings into biolink model?

Attached is a file of all the RO properties that don't biolink-normalize easily. Can we determine whether any of them should be mapped to existing or new biolink model slots?

RO_with_names.txt

[cbizon]

@mbrush @cmungall do you think this is something we could do in the prototype phase?