MOZAIC: Compound Growth via In Silico Reactions and Global Optimization using Conformational Space Annealing

Area of Science:

• Computational chemistry
• Drug discovery

Background:

• Fragment-based drug discovery (FBDD) efficiently explores chemical space using small molecular fragments.
• Computational methods, including AI, are advancing FBDD, but often lack synthetic pathway generation.
• This limitation can hinder the practical synthesis of computationally designed molecules.

Purpose of the Study:

• To introduce MOZAIC, a novel reaction-based fragment-growing framework.
• To address the challenge of generating synthesizable compounds in FBDD.
• To combine in silico reactions with global molecular optimization.

Main Methods:

• MOZAIC employs SMARTS-defined organic reactions to generate compounds, preserving reaction histories.
• Conformational Space Annealing is integrated for global molecular optimization.
• A modular objective function allows for flexible design goals, such as optimizing solubility.

Main Results:

• MOZAIC generated chemically diverse molecules with improved binding affinity, drug-likeness, and synthetic accessibility across benchmark targets.
• The framework demonstrated broad scaffold coverage while maintaining target-directed optimization compared to existing methods.
• Putative synthetic routes were provided for all generated compounds.

Conclusions:

• MOZAIC offers a significant advancement in FBDD by integrating synthesis into the design process.
• The framework successfully generates diverse and optimized molecules with feasible synthetic pathways.
• MOZAIC enhances the efficiency and practicality of computational drug discovery.