Odd results

[patrickkwang]

This is in on the DILI iccees, the query is

{"message":{"query_graph":{
  "edges": {
    "e0": {
      "object": "n1",
      "predicates": [
        "biolink:correlated_with"
      ],
      "subject": "n0"
    }
  },
  "nodes": {
    "n0": {
      "ids": [
        "MONDO:0005359"
      ],
      "is_set": False,
      "name": "drug-induced liver injury"
    },
    "n1": {
      "categories": [
        "biolink:DiseaseOrPhenotypicFeature"
      ],
      "is_set": False,
      "name": "Disease Or Phenotypic Feature"
    }
  }
}
}}

i.e. phenotypic features correlated with DILI.

There are things coming back that I'm not sure are phenotypic features? e.g. sex, racePCA and ethnicity.

Also, sex has a bogus coding: MESH:C110500. The node normalizer claims that this is a chemical substance, and the description from mesh says: "a commercially-available herbomineral formulation; contains Prativisha (Aconitum palmatum), Cowrie bhasma, Shankh bhasma, Maricha (Piper nigrum), Vidanga (Embelia ribes), Triphala & Sunthi (Zingiber officinale); do not confuse with Gas-X"

The identifier for race is also wrong, and the one for ethnicity claims to be 'retired'.

--@cbizon

[karafecho]

I intentionally mapped sex, racePCS, and ethnicity to PhenotypicFeature, realizing that this would be a source of contention, depending on one's perspective. We could change the mapping to something else, such as biolink:BiologicalSex, which likely won't be used by Translator, or we could leave it as is.

WRT the identifier mappings, yes, they are a complete mess. This is something that Priya and I were hoping to resolve pre-relay but were not able. I created a new "ground truth" all_features_search_terms.yaml file pre-relay and Priya used that and Node Norm to generate a new identifiers.yml file. I am now QC'ing the new identifier YAML file and filling in some of the blanks. I hope to complete that task tomorrow. We will then focus on the FHIR mappings YAML file, which is less important than the other two files right now.

Issue to be discussed during 11 am ET 10/6 ICEES+ architecture, design, operations meeting.

[cbizon]

If there's a source of contention, then IMO the right place to handle it is at the biolink level- otherwise different teams will solve the same problem in different ways and none of the solutions will be broadly useful interoperable.

[karafecho]

"Contentious" might have been too strong a word.

The Biolink team is well aware of this issue and others, as they have been working closely with the Clinical Data Committee. The question, in my mind, is whether to map these data elements, and the 'knowledge' they impart, to Biolink or wait until the modeling issues have been resolved.

[cbizon]

Fair enough ... I guess that until we have a good biolink mapping for these things I think they're unlikely to be well used in translator, and probably not worth the trouble to map, but I can see the other side of that.