TRAPI 1.4 validation error when interacts_with predicate is in QEdges or Edges

[colleenXu]

Previously I identified that there was a TRAPI 1.4 validation error whenever interacts_with was the predicate for a QEdge or Edge.

This kind of error can be raised during BTE's creative-mode for chem affects gene because the 4th and 5th templates for both "increases" and "decreases" use interacts_with - so responses can theoretically have Edges with interacts_with. But in practice, it seems uncommon for BTE to get to the 4th template.

Note: the most recent changes to the templates did not adjust / add the interacts_with.

Example responses that raise this error:

• creative-mode:
  • what chem increases activity/abundance of XIST NCBIGene:7503: XISTincreased.txt
  • what chem increases activity/abundance of MALAT1 NCBIGene:378938: MALAT1increased.txt
  • what chem decreases activity/abundance of MAPK8IP3 NCBIGene:23162: MAPK8IP3decreased.txt
• not-creative-mode:
  • Nortriptyline DRUGBANK:DB00540 interacts with what other genes: interacts.txt

The reasoner-validator raises the following:

• for Edge: error.knowledge_graph.edge.predicate.mixin. Note that "errors" are "orange" in the ARAX UI, vs the red for "critical" errors.
• for QEdge: warning.query_graph.edge.predicate.non_canonical. In ARAX UI, responses can have warnings and stay "green"

Snippet of error info for Edges:

{'infores:ddinter -> infores:biothings-ddinter -> infores:service-provider-trapi': {'biolink:interacts_with': [{'edge_id': 'PUBCHEM.COMPOUND:4543[biolink:SmallMolecule]--biolink:interacts_with->PUBCHEM.COMPOUND:24872560[biolink:SmallMolecule]'},
   {'edge_id': 'PUBCHEM.COMPOUND:4543[biolink:SmallMolecule]--biolink:interacts_with->PUBCHEM.COMPOUND:6741[biolink:SmallMolecule]'},
   {'edge_id': 'PUBCHEM.COMPOUND:4543[biolink:SmallMolecule]--biolink:interacts_with->PUBCHEM.COMPOUND:4173[biolink:SmallMolecule]'},

Error info for QEdge:

{'warning.query_graph.edge.predicate.non_canonical': {'global': {'biolink:interacts_with': [{'edge_id': "n0[biolink:SmallMolecule]--['biolink:interacts_with']->n1[biolink:SmallMolecule]"}]}}}

---

These errors are occurring because interacts_with is a mixin and doesn't have canonical_predicate information.

[colleenXu]

I consider this less important and unlikely to cause issues downstream...but maybe it'd be good to ask about this anyways?

• Regarding creative-mode, this could be addressed by replacing interacts_with -> physically_interacts_with for the templates. But if we want to get this in before code-freeze, it has to be done ASAP (@tokebe , thoughts?)
• This originates from us (our x-bte annotations using interacts_with), so we'd be the ones starting data-modeling / reasoner-validator discussions.
• We previously had issues with interacts_with being an abstract term, but I didn't get clear guidance on how to proceed (like what other predicates to use) Issues with abstract terms: interacts_with biolink/biolink-model#1171. The predicate isn't an abstract term anymore.

[colleenXu]

Addressed by replacing interacts_with -> physically_interacts_with in the templates.

Retested here and I confirmed that removal of the templates addresses this issue. Specifically: the TRAPI validation error isn't raised anymore for the response to what chem decreases activity/abundance of MAPK8IP3 NCBIGene:23162: new_MAPK8IP3decreased.txt

As noted in this post, I didn't notice any TRAPI errors when retesting biothings/bte_trapi_query_graph_handler#169 (comment)