TargetSA: adaptive simulated annealing for target-specific drug design

Area of Science:

• Computational chemistry
• Drug discovery
• Artificial intelligence in medicine

Background:

• Target-specific drug design aims to identify compounds with high binding affinity.
• Existing deep generative models face challenges with complex datasets, training, and multi-constraint optimization.

Purpose of the Study:

• To develop a novel framework for target-specific molecular generation and multi-constraint optimization.
• To address limitations of current deep generative models in drug design.

Main Methods:

• Proposed TargetSA framework utilizing adaptive simulated annealing (SA).
• Employed iterative molecular graph editing with insertion, replacement, deletion, and cyclization operations.
• Implemented a reversible sampling strategy to enhance generation quality.

Main Results:

• TargetSA achieved state-of-the-art performance in generating high-affinity molecules (average Vina Dock -9.09).
• Generated molecules maintained desirable chemical properties.
• Framework efficiently explores drug-like chemical space.

Conclusions:

• TargetSA offers an effective solution for target-specific molecular generation.
• The framework successfully integrates multi-constraint optimization into the drug design process.
• Adaptive simulated annealing provides a robust approach for exploring molecular structures.