Create mvp1-template8-service-provider-semmed.json

[mbrush]

adding mvp1 template for bte/service provider semmed ingest - based treats predictions

[mbrush]

Note that this template is essentially identical to template10 defined by tmkp to generate treats predictions from treats or studied or applied to treat edges found in TMKP.

The only difference is the attribute constraint in the tmkp defined template, which I think will be removed.

These templates are identical because the cqs spec does not provide a way define a primary source filter (which is the differentia we need here - semmeddb vs tmkp).

Consider implications of this, and what if anything we need to do to address any issues arising.

The simplest fix for now is to remove the attribute constraint in the tmkp template, and collapse the two templates (template 8 and 10) into one. As defined, this single template should generate all desired treats predictions from both semmed and tmkp content.

But longer term, we should write into the cqs spec a way to constrain/filter based on primary source - as KP / ARA level allow lists are often not sufficient to do this.

[bill-baumgartner]

@mbrush I agree with your comment. In the short term this should replicate results as they were prior to the treats refactor. In the longer term, we likely will want to differentiate based on primary source as you mentioned.

[karafecho]

To clarify, the only reason the allowlist parameter is insufficient is because both templates point to infores:service-provider-trapi. Is that correct? If so, then perhaps the better solution is for infores:service-provider-trapi to expose more granular TRAPI endpoints? FWIW, Multiomics EHR Risk Provider encountered similar issues (see #1).