New research from MIT has heralded a major advance in AI-enabled drug discovery, with the creation of a new AI model that can predict how well drug molecules bind to their protein targets with new levels of speed and accuracy.
The Boltz-2 model was developed by MIT’s Computer Science and Artificial Intelligence Laboratory and the Jameel Clinic, in collaboration with biotech company Recursion.
According to the team, Boltz-2 is the first deep learning model to match the accuracy of intensive physics-based simulations, while running more than 1,000 times faster.
“This performance increase makes Boltz-2 not just a research tool, but a practical engine for real-world drug development,” said Gabriele Corso, an MIT Ph.D. student and one of the lead researchers. “Instead of spending hours simulating the interaction between a single molecule and its target, scientists can now screen vast chemical libraries within the same time frame, enabling early-stage teams to prioritize only the most promising compounds for lab testing.”
The design builds on earlier AI models, such as AlphaFold, which can predict the 3D structures of proteins but couldn’t predict how strongly two molecules could bind together – a critical marker of drug efficacy.
Boltz-2 was designed to meet this gap. Trained on millions of real lab measurements, the AI model can predict binding strength across several benchmarks with “unprecedented accuracy.”
The model builds on the collaborators’ first iteration, Boltz-1, which was designed in 2024. The updated version was retrained using a larger and more diverse dataset, including computer simulations of molecules in motion and synthetic data made of predictions from the earlier version of the model.
“This release is especially significant for the field of small molecule drug discovery, where progress has lagged behind the rapid gains seen in biologics and protein engineering,” said Jameel Clinic researcher Saro Passaro.
“While models like AlphaFold and Boltz-1 allowed a significant leap in the computational design of antibodies and protein-based therapeutics, we have not seen a similar improvement in our ability to screen small molecules, which make up the majority of drugs in the global pipeline,” Passaro added. “Boltz-2 directly addresses this gap by providing accurate binding affinity predictions that can dramatically reduce the cost and time of early-stage screening.”
Boltz-2 is set for general release as fully open source, including model code, weights and training data.