Synthesizability via reward engineering: expanding generative molecular design into synthetic space

Key Points

• The research aims to integrate generative molecular AI with reward functions to create novel molecules that are feasible for synthesis.
• Utilized generative molecular AI techniques to design new molecules.
• Implemented retrosynthetic reward functions to evaluate synthetic feasibility.
• Aligned molecular designs with modern medicinal chemistry practices.
• Produced novel, drug-like molecules optimized for synthesis.
• Ensured compatibility with contemporary parallel synthesis workflows.

Abstract