Understanding Off-Target Effects of Drug Candidates
Case Study 04
A layered assessment combined BindingDB similarity searches with purpose-built target and kinase models to identify potential liabilities and guide follow-up testing.
Challenge
The client needed to understand potential off-target effects across a set of drug candidates and identify which kinases should be prioritized for experimental testing.
What We Did
-
1
Screened for potential off-targets
Used MegaPredict with the BindingDB dataset to identify potential off-targets for the client's molecules.
-
2
Built a focused model for a target of interest
Curated data for one target, built and validated a machine learning model, then used it to score the client's molecules.
-
3
Modeled human kinase activity
Curated human kinase datasets from ChEMBL and generated machine learning models to assess the client molecules across the kinase set.
-
4
Extended the assessment to follow-on work
Additional work included molecule enumeration and property predictions, including cytotoxicity endpoints such as DILI and nephrotoxicity, alongside desired target activity.
The kinase predictions helped identify which kinases to prioritize for testing.