MIT Lab Unveils 2 AI Models to Search 10^20-10^60 Drug Compounds
Updated
Updated · MIT News · May 20
MIT Lab Unveils 2 AI Models to Search 10^20-10^60 Drug Compounds
1 articles · Updated · MIT News · May 20
Connor Coley’s MIT lab has developed two AI systems—ShEPhERD and FlowER—to speed chemical design and small-molecule drug discovery across a chemical space estimated at 10^20 to 10^60 compounds.
ShEPhERD evaluates potential drug molecules by modeling how their 3D shapes interact with target proteins, giving generative systems more medicinal-chemistry intuition; Coley said pharmaceutical companies are already using it in drug hunts.
FlowER predicts reaction products from chemical inputs while enforcing physical constraints such as conservation of mass and the feasibility of intermediate reaction steps, which the lab found improved prediction accuracy.
The work extends Coley’s broader effort since returning to MIT in 2020 to pair AI with chemistry problems, alongside projects in reaction optimization, lab automation and experimental design.
How does AI learn to design drugs for diseases that have long been considered 'undruggable' by human scientists?
As AI accelerates drug discovery, what will prevent these life-saving medicines from becoming prohibitively expensive for patients?
With AI drugs acing early trials, what is the biggest hurdle preventing them from reaching the market and saving lives?
Accelerating Drug Discovery 1,000-Fold: How MIT’s Open-Source AI Models Are Transforming Medicine
Overview
Artificial intelligence is rapidly transforming drug discovery, with MIT leading key innovations such as the Boltz-2 model for molecular modeling and the use of generative AI to design new antibiotics. These breakthroughs are being released as fully open source, making advanced tools accessible to both academic and commercial users. This commitment to openness is democratizing drug discovery, allowing more researchers to participate and accelerating the development of life-saving medicines. By combining powerful, user-friendly AI models with a focus on collaboration and accessibility, the landscape of drug discovery is being reshaped for greater speed, efficiency, and impact.