Improving hERG and Solubility for PROTACs
Generative modeling was adapted to design PROTAC structures and propose strategies to reduce hERG IC50 while improving solubility across multiple PROTACs and a small molecule.
Challenge
The project aimed to propose strategies for reducing hERG IC50 and improving solubility for multiple PROTAC molecules and one small molecule.
What We Did
-
1
Built hERG and solubility models
Curated large datasets for both properties and used them to build machine-learning models.
-
2
Adapted MegaSyn for PROTAC generation
Modified the generative approach to produce PROTAC structures using a database of known linkers.
-
3
Scored and assessed generated molecules
Evaluated generated molecules with the hERG and solubility models, assessed drug-likeness, and applied our own retrosynthesis-scoring approach and similarity searches against SureChEMBL to assess freedom to operate.