ICEES+ / ICEES KG - CAM - AOP Integration

[karafecho]

This issue is intended to stimulate progress on SA3a and DUC2 of our Exposures Provider milestones:

SA3a: Explore and implement technical approaches to integrate ICEES KG KP and CAM/AOP KP, focusing initially on AOPs as identified under DUC2. Develop workflows and TRAPI queries that support the use of CAMs to derive mechanistic insights into exposure-related adverse clinical outcomes derived from ICEES KG KP and other Translator clinical KPs. Expose results as part of the TCDC’s get_creative() ARA queries and other get_creative() ARA queries. Contribute results back to the AOP-Wiki community and the Translator Consortium.

DUC2: In collaboration with the EHLC and AOP-Wiki communities, identify and implement high-impact use-case questions related to AOPs. Use those use-case questions to drive the technical work proposed under SA3a.

As we've discussed, I would like to leverage CAM KP to propose AOPs (or parts thereof) based on ICEES KG observations. Ideally, in addition to the TCDC workflow, we'd start with EHLC's driving use case on exposure to particulate matter and asthma-related outcomes such as hospitalizations or the need for prednisone (typically prescribed to treat severe asthma).

Below is an ICEES+ query looking at the relationship between particulate matter exposure and prednisone:

curl -X 'POST' \
  'https://icees-asthma.renci.org/patient/cohort/COHORT%3A1/feature_association?year=2010' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "feature_a": {
    "PrednisoneRx": {
      "operator": "=",
      "value": "1"
    }
  },
  "feature_b": {
    "AvgDailyPM2.5Exposure_StudyMax": {
      "operator": "<",
      "value": "3"
    }
  }
}'

Below is an ICEES KG TRAPI query that returns 'particulate matter' in the answer set for ChemicalEntity - RWE - Prednisone (note that EnvironmentalExposure - RWE - Prednisone should work, too):

`{
  "message": {
    "query_graph": {
      "nodes": {
        "n0": {
          "categories": [
            "biolink:ChemicalEntity"
          ]
        },
        "n1": {
          "ids": [
            "PUBCHEM.COMPOUND:5865"
          ],
          "categories": [
            "biolink:ChemicalEntity"
          ]
        }
      },
      "edges": {
        "e0": {
          "subject": "n0",
          "object": "n1",
          "predicates": [
            "biolink:has_real_world_evidence_of_association_with"
          ]
        }
      }
    }
  },
  "workflow": [
    {
      "id": "lookup"
    }
  ]
}`

Can we use CAM to fill in these clinical observations with mechanistic understanding related to, say, genes and/or biological processes?

Note that this issue has relevance to both Translator and the ROBOKOP U24.

Note also that JSON responses to both queries above can be found here.

[gaurav]

Hi Kara! Here's what dev-CAM-KP currently knows about chemicals related to prednisone (which we refer to as CHEBI:8382). I had to add another node in there because CAM-KP models this as e.g. CHEBI:8382 -participates-in-> (process) <-participates-in- CHEBI:27208. There are three problems with this at the moment:

• That intermediate process just shows up as BFO:0000015 (process); I think it should be something more specific. This is just how CTD is modeled right now.
• We shouldn't need to go through the intermediate process to figure out that these two chemicals are related to each other -- we should improve that part of the reasoning in cam-pipeline.
  • This might be because the two chemicals are being used together rather than affecting each other -- see http://ctdbase.org/detail.go?type=relationship&ixnId=2764668
• All the results appear to be duplicated for some reason.

[capasfield]

https://icees-asthma.renci.org/apidocs#/ – ICEES +
https://automat.renci.org/#/ - ICEES KG (development)