MIT and Recursion Collaborate on Innovative AI Tool for Drug Discovery

A team led by MIT professor Regina Barzilay has developed Boltz-2, an innovative algorithm combining protein folding and small-molecule binding affinity predictions. This AI model, born from collaboration with Recursion Pharmaceuticals, addresses a long-standing challenge in the field. Barzilay notes that this advancement required novel machine learning approaches. Boltz-2 achieves human-level accuracy comparable to traditional computational chemistry methods, but with significantly enhanced speed and cost-effectiveness. While the model is publicly available for academic and commercial use, it has yet to be peer-reviewed. Trained on diverse datasets, including internal data from Recursion, Boltz-2 has already outperformed its predecessor, Boltz-1, particularly in areas like virtual screening and lead optimization. Recursion’s chief R&D officer, Najat Kahn, underscores the importance of ongoing experimental validation to further enhance the algorithm’s capabilities.

Source link