Chai Discovery co-founder Matthew McPartlon walks through whether AI can design antibody therapeutics computationally, from structure and sequence generation through experimental validation. Speaking at Greylock and Altara’s AI for Science Symposium, he explains how advances in protein folding enabled a new generation of design systems and how Chai pairs generation with model-based filtering.
The headline number is a jump from roughly 0.1% prior state-of-the-art hit rates to 18.1%, alongside close agreement between predicted and experimentally measured structures. McPartlon also makes the case that wet-lab validation still matters even as computational design improves, and covers the open-source versus proprietary split in antibody models.
