MIT and Recursion Release Boltz-2: Next Generation AI Model to Predict Binding Affinity at Unprecedented Speed, Scale, and Accuracy

Core Insights - The article discusses the launch of Boltz-2, an open-source biomolecular foundation model developed by MIT and Recursion, which significantly improves the accuracy and speed of predicting molecular binding affinities and structures [1][2][5]. Company and Industry Overview - Boltz-2 is a pioneering model that combines structure and binding affinity prediction, achieving near-physics-based accuracy while being over 1,000 times faster than traditional methods [5][6]. - The model is designed to enhance drug discovery processes by allowing researchers to select promising molecules more effectively, thereby improving the success rates of R&D programs [3][4]. - The open-source nature of Boltz-2, including its training code, enables scientists to customize the model for specific molecules, facilitating broader applications in both academic and commercial settings [2][4]. - Recursion, the TechBio company behind Boltz-2, utilizes advanced machine learning algorithms and operates one of the world's most powerful supercomputers, BioHive-2, to support its mission of decoding biology for improved healthcare outcomes [7][8]. - The development of Boltz-2 involved collaboration between MIT's academic expertise and Recursion's AI capabilities, highlighting the importance of partnerships in advancing biotechnological innovations [3][5].