Researchers from the Indian Institute of Technology Madras (IIT Madras) and Ohio State University have developed an advanced artificial intelligence framework that replicates real-world chemical synthesis to accelerate drug discovery. The system, named PURE (Policy-guided Unbiased Representations for Structure-Constrained Molecular Generation), generates drug-like molecules that are easier to synthesize in laboratories—solving a long-standing challenge in computational drug design.

Unlike traditional AI models that rely on fixed scoring systems, PURE uses reinforcement learning and self-supervised learning to understand how molecules transform during chemical reactions. This allows the system to propose both viable molecular structures and realistic synthetic routes.

The framework demonstrated superior results across several key performance metrics, including drug-likeness, dopamine receptor activity, and solubility.

Professor B. Ravindran from IIT Madras explained that PURE represents a shift from optimizing only numerical properties to learning the actual chemistry behind molecular transformations.

By integrating chemical synthesis knowledge into molecular generation, this innovation has the potential to drastically reduce time and cost in early-stage drug development, helping pharmaceutical researchers design and produce new therapeutic compounds more efficiently.