Enabling Synthetically Feasible Molecular Editing in Drug Discovery via Reaction-Regulated Graph-Based Genetic...

Area of Science:

• Computational chemistry
• Artificial intelligence in drug discovery
• Molecular modeling

Background:

• Machine learning generative models accelerate chemical space exploration.
• Current models often neglect synthetic feasibility, limiting practical applications.
• Integrating theoretical and experimental perspectives remains a challenge.

Purpose of the Study:

• To develop a novel method for synthetically feasible molecular design.
• To enable site-selective molecular modifications while preserving scaffold integrity.
• To improve the exploration of complex chemical space for molecular editing.

Main Methods:

• Proposed a reaction-regulated graph-based genetic algorithm (R²GB-GA).
• Embedded domain-specific reaction rules into the evolutionary algorithm.
• Utilized fragment-based and pathway-based scoring for evaluation.

Main Results:

• R²GB-GA generates more synthetically accessible molecules compared to conventional methods.
• The method successfully performs site-selective modifications on molecular scaffolds.
• Demonstrated practical applicability in designing ligands for heat shock protein 90 inhibition.

Conclusions:

• The reaction-based framework enhances molecular design by ensuring synthetic feasibility.
• R²GB-GA offers a robust approach for molecular editing and exploration.
• This method has broad applicability in drug discovery, including late-stage functionalization.