CTD should not have "Treats"

[cbizon]

The question is "What treats polymyositis?" in PROD
PK = 7aa99faa-37e6-4a63-a604-7d3e73b908d3

That treats edge seems to be coming from CTD via BTE:

CTD doesn't really report "Treats" as we understand it. It looks like it's coming via molepro. And I'm seeing the same edge in CI.

[cbizon]

@mbrush

[vdancik]

@cbizon CTD provides "therapeutic" flag for chemicals:

[cbizon]

That's true, but I don't think that CTD's "Therapeutic" rises to the level of treats in biolink:

‘ Treats’ edges should be asserted (knowledge_level: assertion) only in cases where there
is strong supporting evidence - i.e. in some population(s) the intervention is approved for
the condition, passed phase 3 or in phase 4 trials for the condition, or is an otherwise
established treatment in the medical community (e.g. a widely-accepted or formally recommended
off-label use). In the absence of such evidence, weaker predicates should be used in
asserted edges (e.g. ‘in clinical trials for’ or ‘beneficial in models of’).

I think a mapping of treats_or_studied_or_applied_to_treat would be more appropriate (and is how we ingest CTD in robokop).

[mbrush]

That was also my recommendation based on what you told me about the provenance of CTD 'T' annotations, and the level of confidence we have that they represent true 'treatments' that fit our definition and implementation guidance around the 'treats' predicate. @vdancik if you think differently (i.e. that CTD 'T' annotations are strong enough to warrant a 'treats' assertion), happy to have a conversation about this.