Google DeepMind · De Novo Enzyme Design

AlphaProtein Novo

Google DeepMind's open-sourced pipeline for designing new enzymes, announced 5 October with a bioRxiv preprint written with Frances Arnold's lab at Caltech and Peng Liu's group at Pittsburgh. A diffusion model designs protein structure and amino-acid sequence together around a given catalytic motif and ligand, optionally followed by LigandMPNN sequence redesign and AlphaFold 3 structure filtering. The preprint reports more than 5,600 designs tested across five reactions with hit rates up to 80%, nitrene transferases that make piperidine at 99:1 regioselectivity, and an esterase that breaks down the plasticizer DEHP and is 14 times more active at 90°C than at room temperature. Code is Apache-2.0; model weights are a separate non-commercial download.