Approach for the Design of Covalent Protein Kinase Inhibitors via Focused Deep Generative Modeling.

Area of Science:

• Computational chemistry
• Drug discovery
• Machine learning

Background:

• Covalent inhibitors are gaining importance in drug discovery, especially for protein kinases.
• Computational design of covalent inhibitors remains underexplored.

Purpose of the Study:

• To develop a computational approach for designing covalent protein kinase inhibitors.
• To apply this approach to Bruton's tyrosine kinase (BTK) as an exemplary target.

Main Methods:

• Combining fragment-based design with deep generative modeling.
• Augmenting the generative model with 3D pharmacophore screening.
• Learning from kinome-relevant chemical space for systematic inhibitor design.

Main Results:

• Generated novel covalent inhibitor candidates for BTK, including known compounds and new structures.
• Demonstrated the approach's ability to generate chemically reactive compounds for covalent modification.
• Required minimal target-specific information for effective design.

Conclusions:

• The developed computational approach is effective for designing covalent kinase inhibitors.
• This method integrates knowledge-based design with deep learning, offering a chemically intuitive strategy.
• The approach is readily applicable to other kinase targets in drug discovery.