DeepMind's AlphaProtein Novo designs de-novo enzymes with SOTA activity on benchmark reactions
pushmeet · x · 2026-10-05
Google DeepMind's team shares a preprint on AlphaProtein Novo, a system for generative de-novo enzyme design — the next step after AlphaFold, protein language models, and AlphaProteo. The system produces new-to-nature enzymes with state-of-the-art activity on 2 benchmark reactions, plus custom enzymes to synthesize the pharmaceutical motif piperidine and to degrade the environmental plasticizer toxin DEHP. The authors note much work remains.
Related event: DeepMind Unveils AlphaProtein Novo for De Novo Enzyme Design(7 posts)→
More from AGI Musings
- Consciousness scholar Anil Seth: Life is a prerequisite for conscious AI — anilkseth · 2026-10-06
- Economist Voth: Scarce labor and high wages can boost the economy — sebkrier · 2026-10-06
- AI researcher asks: what task can't a tool-using LLM do by 2030? — Afinetheorem · 2026-10-06
- Self-replicating robot factories: AI agents coordinating modular robot cells — ihorbeaver · 2026-10-06
- Anil Seth's new essay argues conscious AI is a mythology — haydendevs · 2026-10-06
- Altman: the world should accept some bad things happening for AI's benefits — Afinetheorem · 2026-10-06