In a significant advancement for pharmaceutical research, scientists at The Ohio State University have created an artificial intelligence system that could revolutionize how new medications are developed.
The novel generative AI model, called DiffSMol, was developed by a team led by Professor Xia Ning from the university's departments of biomedical informatics and computer science and engineering. DiffSMol works by analyzing the shapes of known ligands – molecules that bind to protein targets – and using these shapes as conditions to generate entirely new 3D molecules with enhanced binding properties.
"By using well-known shapes as a condition, we can train our model to generate novel molecules with similar shapes that don't exist in previous chemical databases," explained Ning. The system's effectiveness is remarkable – when creating molecules with the potential to accelerate drug development, DiffSMol achieved a 61.4% success rate, dramatically outperforming previous research attempts that managed only about 12% success.
The researchers demonstrated DiffSMol's capabilities through case studies on molecules targeting cyclin-dependent kinase 6 (CDK6), which can regulate cell cycles and disrupt cancer growth, and neprilysin (NEP), used in therapies aimed at slowing Alzheimer's progression. Results showed the AI-generated molecules would likely be highly effective, with DiffSMol outperforming baseline methods in binding affinities by 13.2%, and by 17.7% when combined with shape guidance.
This breakthrough comes as the FDA establishes new regulatory frameworks for AI in drug development. In January 2025, the agency released draft guidance titled "Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products," providing recommendations on using AI to support regulatory decisions regarding drug safety, effectiveness, and quality.
While traditional drug development typically takes about a decade from discovery to market, AI-powered approaches like DiffSMol could significantly compress this timeline. The research team has made DiffSMol's code available to other scientists, though they acknowledge current limitations – the system can only generate new molecules based on shapes of previously known ligands, a constraint they hope to overcome in future work.