The Indian Institute of Technology Madras (IIT Madras) has announced a breakthrough in artificial intelligence-driven drug discovery with the launch of ‘PURE’ (Policy-guided Unbiased REpresentations for Structure-Constrained Molecular Generation), an innovative AI framework designed to rapidly generate drug-like molecules that are more feasible to synthesise in real-world laboratories.
Developed jointly by researchers from IIT Madras’ Robert Bosch Centre for Data Science and AI, the Wadhwani School of Data Science and AI (WSAI), and Ohio State University in the United States, the framework aims to dramatically reduce the time and cost involved in early-stage drug development, currently a process that can take over a decade and cost billions of dollars.
According to IIT Madras, PURE stands apart from conventional AI molecule-generation tools that depend heavily on rigid scoring metrics or statistical optimisation. Instead, it leverages reinforcement learning to simulate how molecules transform through real chemical reactions, enabling AI systems to reason about synthesis steps much like a human chemist.
Prof B Ravindran, Head of WSAI, IIT Madras, explained that what makes PURE unique is the way it uses reinforcement learning not just to optimise certain metrics, but to understand how molecules evolve. By treating chemical design as a sequence of actions guided by real reaction rules, PURE brings research closer to AI systems that can think like chemists.
The framework was tested on several benchmark datasets, including QED (drug-likeness), DRD2 (dopamine receptor activity), and solubility tests, showing superior results in generating molecules that are both effective and synthesisable.
Prof Karthik Raman, WSAI, IIT Madras, highlighted that PURE uses a reaction rule-based approach to explore chemical space without bias, ensuring the molecules it proposes are realistically synthesisable.
Prof Srinivasan Parthasarathy from Ohio State University noted that PURE offers game-changing benefits for pharmaceutical research, particularly in finding alternatives to drugs facing resistance and hepatotoxicity. He added that it blends cutting-edge self-supervised learning with policy-based reinforcement learning, using template-driven molecular simulations to explore chemical possibilities.
Beyond drug discovery, the PURE framework also shows promise for new material discovery, providing a foundation for future advances in computational chemistry and materials science.
The research findings have been published in the peer-reviewed Journal of Cheminformatics, underscoring the growing impact of AI and data science in transforming how new drugs and materials are discovered.
