MyChem drugcentral bioactivity: what to do when action_type is missing

[colleenXu]

MyChem's drugcentral.bioactivity info is a resource for detailed chem-gene relationship information. We currently use the field action_type to help map the data to biolink predicate/qualifier info (ex: the value "AGONIST" -> affects / causes increased activity using agonism). The action_type field seems to correspond to the "Action" column in the DrugCentral webpages (ex: compare the drugcentral.bioactivity data for bupropion in MyChem vs DrugCentral's webpage).

However, as you can see in the bupropion links, there's a LOT of data that doesn't have an action_type value: the field is missing in the MyChem API data and is blank on the DrugCentral webpage. I'm not sure how to quantify it in MyChem, but I know there's 763 / 3225 chemicals in MyChem's drugcentral bioactivity data that are MISSING action_type for ALL of their gene relationships.

In order to make this data available to Translator, there's two things we'd need to figure out:

• we're not sure how to model it with biolink-model: what predicate to use? Affects?
• we're not sure how to query for it in the BioThings API. It's complicated when chemicals have a mix of data with the action_type and without it. -> Trying to ask Chunlei and co.

[colleenXu]

Thanks to @DylanWelzel, there is a way to query for these in BTE! use jmespath=drugcentral.bioactivity|[?!action_type]

Query from chem ➡️ gene (forward, can be batch):
If you take out the jmespath param, the response will have 2 bioactivity objects with action_type INHIBITOR.

curl --location --globoff 'https://mychem.info/v1/query?size=1000&fields=drugcentral.bioactivity%2Cdrugcentral.xrefs.umlscui%2Cdrugcentral.synonyms&jmespath_exclude_empty=true&always_list=drugcentral.bioactivity&jmespath=drugcentral.bioactivity%7C[%3F!action_type]' \
--header 'Content-Type: application/json' \
--data '{
    "q": ["C0028365"],
    "scopes": "drugcentral.xrefs.umlscui"
}'

Query from gene ➡️ chem (reverse, can't be batch):

curl --location --globoff 'https://mychem.info/v1/query?size=1000&fields=drugcentral.bioactivity%2Cdrugcentral.xrefs.umlscui%2Cdrugcentral.synonyms&jmespath_exclude_empty=true&always_list=drugcentral.bioactivity&jmespath=drugcentral.bioactivity%7C[%3F!action_type%20%20%26%26%20length(uniprot[%3Funiprot_id%3D%3D%27P47869%27])%20%3E%20%600%60]' \
--header 'Content-Type: application/json' \
--data '{
    "q": ["P47869"],
    "scopes": "drugcentral.bioactivity.uniprot.uniprot_id"
}'

[colleenXu]

Since it's possible to query for this data in the BioThings API, I went ahead and made x-bte operations for it (two commits). I also added these changes to the remove-clinical-trials branch so all instances of BTE/Service Provider are using it starting NOW (they are currently using overrides to that branch, see #861 (comment)).

I decided to use the predicate interacts_with for now (no qualifiers) because:

• "bioactivities" appear to be "drug-target potency pairs", where the supporting data is sometimes a Kd (binding constant).
• I think this is at least an assertion that the chemical is interacting with the protein (gene's product)
• If a bioactivity is supported with a Kd (or no measurement data), I'm not sure if the drug really creates an effect. (does binding, even tight and specific binding, necessarily mean there's an effect?) I feel like other assays need to be done to confirm an effect.

---

This is my understanding of what DrugCentral's "bioactivities" are (bold emphasis mine):

• it's pharmacodynamics, which is the study of "a drug's molecular, biochemical, and physiologic effects or actions". Chemicals interact with biological structures (ex: cell membrane), target-proteins, target-molecules, etc. to produce these effects. These interactions and effects are measured using biochemical assays.
• DrugCentral compiles "unique drug-target potency pairs", preferring "binding (Kd) and inhibition (Ki) constants over other activity measurements types such as IC50" (Ref: "Bioactivity profiles" section of original DrugCentral paper. Also note def of drug potency)

[colleenXu]

Going to close this issue for now.

Comments can be added/issue can be reopened if Translator data-modeling or chem-specialist folks have thoughts on how to handle/map this data.