Issues with CTD modeling

[gaurav]

We have run into some problems with modeling CTD data that makes it hard to get useful information from CTD models. These were discovered as part of #599 and are visible on the Google CoLab nodebook that goes with it.

If you go to https://cam-kp-api-dev.renci.org/1.3.0/docs/index.html?url=docs.yaml#/default/postQuery and run the following TRAPI query for things related to prednisone (CHEBI:8382):

{
  "message": {
    "query_graph": {
      "nodes": {
        "n0": {
          "categories": [
            "biolink:NamedThing"
          ]
        },
        "n1": {
          "ids": [
            "CHEBI:8382"
          ]
        }
      },
      "edges": {
        "e0": {
          "subject": "n0",
          "object": "n1"
        }
      }
    }
  }
}

We only get back the following relationships:

• (http://purl.obolibrary.org/obo/RO_0000056 ("participates in"), http://purl.obolibrary.org/obo/RO_0002328 ("functionally related to"), http://purl.obolibrary.org/obo/RO_0002352 ("input of"))
  • http://purl.obolibrary.org/obo/BFO_0000015 ("process")
• (http://purl.obolibrary.org/obo/RO_0000056 ("participates in"))
  • http://purl.obolibrary.org/obo/GO_0042221 ("response to chemical")
  • http://purl.obolibrary.org/obo/GO_0046903 ("secretion")
  • http://purl.obolibrary.org/obo/BFO_0000003 ("occurrent")

We can get additional data by doing a two-hop instead of a one-hop:

{
  "message": {
    "query_graph": {
      "nodes": {
        "n0": {
          "ids": [
            "CHEBI:8382"
          ]
        },
        "n1": {
          "categories": [
            "biolink:NamedThing"
          ]
        },
        "n2": {
          "categories": [
            "biolink:NamedThing"
          ]
        }
      },
      "edges": {
        "e0": {
          "subject": "n0",
          "object": "n1"
        },
        "e1": {
          "subject": "n1",
          "object": "n2"
        }
      }
    }
  }
}

This mostly links to other processes, HOWEVER, we do get back http://id.nlm.nih.gov/mesh/D000077154 ("Rosiglitazone"), which hasn't been normalized, so it can't be connected to other things in TRAPI.