Predicting Off-Target Liabilities for a PROTAC Molecule
← Case Studies
Case Study 12
Predicting Off-Target Liabilities for a PROTAC Molecule
MegaPredict BindingDB, a GraphSage model trained on 475 human kinases, and models for several immunology targets were used to assess a client-provided PROTAC structure.
Challenge
The client provided the structure of a PROTAC molecule for off-target assessment.
What We Did
-
1
Queried MegaPredict BindingDB
CPI queried the similarity of the client molecule to millions of molecules in MegaPredict BindingDB and assessed the top-scoring targets.
-
2
Assessed kinase hits
Because several top-scoring hits were kinases, CPI applied a GraphSage kinase model trained on 475 human kinases to score the client molecule.
-
3
Developed models for immunology targets
Built additional models for several immunology targets and used them to score the client's molecule.