"max predicate approach" for refactoring "treats"

[sierra-moxon]

Treats refactoring proposal, summarized here: https://docs.google.com/document/d/1ff_-SYdixIQrPStKkyUeVfMnLHkvuDziCnkTNH-y8-Y/edit#

Contributing members:
Matt Brush @mbrush
Mike Bada @mikebada
Sandrine Muller @sandrine-m
Vlado Dansik @vdancik
Kamileh Narsinh @GitHubbit

and others in the predicates WG.

[mbrush]

I just updated the 'treats' definition to reflect input form recent conversations with KPs/ARAs. If now reads as follows:

  Holds between an intervention (substance, procedure, or activity) and a medical condition
  (disease or phenotypic feature), and states that the intervention is able to ameliorate, 
  stabilize, or cure the condition or delay, prevent, or reduce the risk of it manifesting 
  in the first place. 
  ‘Treats’ edges should be asserted (knowledge_level: assertion) only in cases where there 
  is strong supporting evidence - i.e. the intervention is approved or in phase 4 trials for 
  the condition, or is an otherwise established treatment in the medical community (e.g. 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’). ‘Treats’ edges based on weaker or indirect forms of evidence can however 
  be created as predictions (knowledge_level: prediction) and should point to the more 
  foundational asserted edges that support them.

My edits introduced back the term 'intervention as subsuming substances, procedures, or activities in the first paragraph. And added the parenthetical "(e.g. widely- accepted or formally recommended off-label use) to the second paragraph - to call out specific example of an established treatment.