Invalid predicates

[kevinschaper]

I'm stripping off invalid predicates in my phenio filtering in monarch-ingest, and found that I have 5128244 edges with predicates that aren't in biolink model, the categories are: 'biolink:category', 'biolink:inverseOf', 'biolink:subPropertyOf'

I'm not sure if these are edges that we want to keep and need to fix the predicates, or if these are just properties that are staying as edges in the conversion.

[matentzn]

What do you mean by "valid"?

category exists: https://github.com/biolink/biolink-model/blob/master/biolink-model.yaml#L507

We should:

• not filter out edges with invalid predicates, but include edges with valid ones (assuming that 1 edge can have more than one predicate)
• Figure out what the deal is with subPropertyOf which is probably added by KGX - and should probably be added to biolink (instead of this horrible mapping)

[caufieldjh]

The majority (maybe even all!) of those 'biolink:category' nodes are leftovers from category mapping in the phenio build and serve no purpose in the final graph. It's fully redundant when the categories are also present as node properties. I've added a step to remove these edges during KG-Phenio assembly but, embarrassingly, that step doesn't work correctly.

[caufieldjh]

The biolink:category and biolink:inverseOf nodes/edges are gone now as of KG build 20241203.
The biolink:subPropertyOf edges are still present as I'm still unclear on how it's getting there.