International collaboration spurs AI-powered drug discovery tool

An error occurred while preparing your download

International collaboration spurs AI-powered drug discovery tool

System simulates step-by-step molecular changes based on real chemical reactions

Researchers from The Ohio State University and the Indian Institute of Technology Madras have developed an artificial intelligence framework to rapidly generate drug-like molecules that are easier to synthesize in real-world laboratory settings.

The new system, called PURE (Policy-guided Unbiased REpresentations for Structure-Constrained Molecular Generation), promises to significantly cut down the early-stage timelines of drug development - currently a billion-dollar, decade-long process - and could play a crucial role in addressing drug resistance in cancer and infectious diseases. It stands apart from existing molecule-generation AI tools that rely on rigid scoring mechanisms or statistical optimization.

PURE draws inspiration from how drugs are actually synthesized in labs, simulating step-by-step molecular changes using templates derived from real chemical reactions. By blending self-supervised learning - which lets the model learn patterns from data without labels - with a policy-based reinforcement learning setup, it explores the chemical landscape more naturally.

One of the biggest problems in AI-driven drug discovery is that most AI-generated molecules look promising on a computer but are nearly impossible to synthesize in reality. PURE solves this.

"This new framework offers game-changing benefits for early-stage pharmaceutical research, with the capability to identify alternative, more effective drug candidates in the face of resistance and hepatotoxicity," explained Ohio State Computer Science and Engineering Professor Srinivasan Parthasarathy. "It blends cutting-edge self-supervised learning with policy-based reinforcement learning, using template-driven molecular simulations to navigate the discrete molecular search space while mitigating metric leakage.

"In addition to drug discovery, PURE provides a promising foundation for accelerating the discovery of new materials, an important future research direction."

PURE was evaluated on widely accepted molecule-generation benchmarks, including QED (drug-likeness), DRD2 (dopamine receptor activity) and solubility tests. It delivered more diverse and original molecules and generated possible synthetic routes without ever being trained on those scoring metrics. This makes PURE a general-purpose AI engine for molecular discovery, capable of working across multiple disease and property objectives using a single trained model.

The findings were published in the Journal of Cheminformatics.

Parthasarathy's collaborators include B. Ravindran, Karthik Raman, Abhor Gupta, Barathi Lenin and Rohit Batra from IIT Madras and recent Ohio State PhD graduate Sean Current.

"What's unique about PURE is the way it uses reinforcement learning, not just to optimize specific metrics, but to learn how molecules transform," said Ravindran. "By treating chemical design as a sequence of actions guided by real reaction rules, PURE moves us closer to AI systems that can reason through synthesis steps much like a chemist would."

In addition to drug discovery, Parthasarathy said the PURE framework also provides a promising foundation for accelerating the discovery of new materials.

More Ohio State News