Design of highly affine and specific protein binders with AlphaProteo

ABSTRACT

Designing novel proteins that bind tightly and specifically to disease-relevant targets is a fundamental challenge in drug discovery and synthetic biology. We introduce AlphaProteo, an integrated machine learning system that designs high-affinity protein binders for diverse target structures. Leveraging the geometric and evolutionary priors learned by AlphaFold, AlphaProteo employs a diffusion-based generative model to sculpt novel binding interfaces. Across rigorous wet-lab validations, AlphaProteo successfully designed binders for several high-value therapeutic targets—including VEGF-A and viral spike proteins—achieving binding affinities in the picomolar to nanomolar range with unprecedented experimental success rates, significantly outperforming traditional computational design methods.