Remove mapping between drug warning and MedDRA

[ireneisdoomed]

Remove look up table between adverse effect and MedDRA code in the ETL step that processes drug warnings (and other downstream tasks).

Background

Now that we have most of the drug warnings mapped to an EFO code, the MedDRA codes we have linked to the toxicity classes are of less value. I suggest removing this specific piece of code from the ETL as it has lost its usefulness over time. The associated dependency is costly to maintain while providing little return. For example, the data has changed in the latest ChEMBL release (toxicity classes are now lower cased), therefore the MedDRA codes are always null (https://github.com/opentargets/issues/issues/3002)

Tasks

• [x] Update https://github.com/opentargets/platform-etl-backend/blob/5be39b55584d48472b3ffb10b2cfb7feb739b1f4/src/main/scala/io/opentargets/etl/backend/drug/DrugWarning.scala#L21 to remove the LUT
• [x] Remove meddraSocCode from the drug.drugWarnings API endpoint
• [X] Remove the MedDRA SOC Code table from the UI

Acceptance tests

How do we know the task is complete?

1. The schema of the drugWarnings dataset does not have a field called `meddraSocCode``
2. The API endpoint drug.drugWarnings.meddraSocCode no longer exists
3. CHEMBL8 does not show any reference to MedDRA in the drug warnings component

[mbdebian]

@jdhayhurst , @ireneisdoomed is going to be around for the next two weeks, would you mind checking with her whether this is issue is still relevant?

[jdhayhurst]

@ireneisdoomed just checking with you that this is still required? It looks like if we remove the meddra LUT step, we're effectively reading in the chembl drug warning data (which comes from the chembl elastic), renaming the fields and then writing it out without any transformations. Just wanted to check that that sounds correct.

[jdhayhurst]

@ireneisdoomed I removed the meddra LUT step here - this removes the meddraSocCode from the drugWarning data. I have also removed the meddraSocCode from drugWarning data model in the API and was about to make the PR when I noticed that you actually said that the meddraCode field should be removed from drug.adverseEvents. This field comes from somewhere else entirely so I wanted to be sure it's right one 😅 . The meddraCode field comes from the OpenFDA step in the ETL. The meddraCode field is ultimately derived from these sources. Please could you confirm that you'd like this meddraCode field to be removed? Thanks!

[ireneisdoomed]

@jdhayhurst Thank you for the changes in https://github.com/opentargets/platform-etl-backend/pull/323 It looks good to me, however it's easier to QC the data next time this is run to make sure we are not missing anything. The API field to be removed is indeed drugWarning.meddraSocCode, not the pharmacovigilance one. Thanks again!

[jdhayhurst]

Keeping this open (the pr's auto closed it) so we can test in dev platform 😉

[ireneisdoomed]

These changes have implications in the FE that we have to action to close the issue. Looking at the acceptance tests:

1. drugWarnings has been removed from the data ✅
2. drug.drugWarnings.meddraSocCode is no longer a field in the API ✅
   • We need to update the example query in the API page https://platform.dev.opentargets.xyz/api. It might not be related, but the chemblId env variable is incorrectly defined for the drug endpoint

3. We are missing the drug warnings widget for ciprofloxacine because the FE hasn't been updated yet ❌

Thank you!

cc @prashantuniyal02 @carcruz