Were the openfda_adverse_target_reactions and ChEMBL evidence datasets discontinued in 26.06?

Technical Support
, ,
1

Hello Open Targets team,

We are updating our data pipeline from Open Targets Platform release 25.09 to release 26.06.

We could not find the following datasets in the 26.06 downloads:

  • openfda_adverse_target_reactions
  • evidence/sourceId=chembl

They appear to be absent from the relevant output, intermediate, and excluded directories.

Could you please confirm:

  1. Were these datasets intentionally discontinued in release 26.06?
  2. If so, is there a replacement dataset, or another recommended source from which we can obtain the equivalent OpenFDA target-level adverse-reaction data and ChEMBL evidence data?

We are especially concerned about the openfda_adverse_target_reactions, as we have not identified a clear replacement for the target-level adverse-reaction information.

Thank you for your help.

2

Hi @spletsx and welcome to our Community!

  1. The openfda_adverse_target_reactions has been discontinued. On 26.06, we concluded the migration of our so-called Scala ETL pipelines into Python. Because the dataset wasn’t really used anywhere in the Platform nor in our internal analyses, we made the decision to discontinue it for the time being. We have started a Target Safety project that aims to use ADRs as evidence to understand target safety.
    On the other hand, the ChEMBL target/disease has been discontinued, yes, but in favour of an analogous dataset that is more up-to-date and pulls information from other databased like TTD on top of the data from ClinicalTrials.gov. This dataset is now called evidence_clinical_precedence. Please let us know if you have any feedback on this space.
  2. For the OpenFDA target safety annotations, I’d suggest taking the drug level annotation and overlapping it with the Drug Mechanism of Action dataset. This is a more naive approach than the previous dataset, because you’re not really doing association testing between targets and adverse reactions, but in essence you end up with a table linking adverse reactions to targets that might be of use. You could tweak our OpenFDA script to run Monte Carlo at the target level, but the amount of input data required is not negligible.

I hope this is useful. Really sorry for the inconvenience, we were not really aware of any users consuming this data. I hope you can find a reasonable workaround!

Let me know if you have any other comments.
Best,
Irene