DeepMind Unveils AlphaProtein Novo for De Novo Enzyme Design
Google DeepMind has released a preprint introducing AlphaProtein Novo, a new de novo enzyme design system, developed in collaboration with 2018 Nobel Prize in Chemistry laureate Frances Arnold and Peng Liu's research group. The pipeline targets small-molecule enzyme design, generating enzymes that don't exist in nature on standard model reactions, with validation completed in DeepMind's own wet lab. The current takeaway: the system has produced competitive catalytic designs across multiple real-world scenarios, marking another milestone as AI protein design moves from structure prediction toward function creation.
Confirmed
- AlphaProtein Novo was released as a DeepMind preprint, integrating capabilities from AlphaFold (structure prediction), protein language models (function understanding), and AlphaProteo (de novo binder design).
- Collaborators include the lab of Frances Arnold (2018 Nobel Prize in Chemistry laureate) and Peng Liu's research group.
- It achieved "novel-scaffold" designs on 2 benchmark reactions.
- The system was first developed on several standard model reactions; the team validated the designs in its own wet lab, then proactively reached out to biologists to expand application scenarios.
- Real-world use cases include the synthesis mechanism of novel piperidines (a common N-heterocyclic scaffold in drugs) and degradation of the toxic pollutant DEHP; the team designed novel enzymes with catalytic performance surpassing natural enzymes.
- According to a repost, initial designs yielded only 0.1%, with performance improving significantly after one round of design optimization.
Why it matters
- Enzymes are central tools in synthetic chemistry and environmental degradation. De novo design of enzymes that catalyze custom reactions means AI is no longer limited to predicting existing proteins—it can directly create functional molecules.
- Partnering with a Nobel-caliber experimental team and validating in a wet-lab closed loop strengthens the reliability of the results and provides a direct template for bringing AI design to industrial applications (drug-scaffold synthesis, pollutant remediation).
2026-10-05 ~ 2026-10-06 · 7 related posts
Primary sources
- [source] AlphaProtein Novo: AI pipeline hits SoTA in de novo enzyme design — jueseph · 2026-10-05
- [source] DeepMind's AlphaProtein Novo designs de-novo enzymes with SOTA activity on benchmark reactions — pushmeet · 2026-10-05
- DeepMind releases AlphaProtein Novo pipeline for small-molecule enzyme design — BrianHie · 2026-10-05
- AlphaProtein Novo designs novel enzymes that outperform natural starting points — snikolov · 2026-10-05
- [source] AlphaProtein Novo pipeline hits SoTA in de novo enzyme design, synthesizes piperidine motif — DaveJuergens · 2026-10-06
- DeepMind's AI designs enzymes from scratch: 99x drug building block yield, plastic-eating at 90°C — 141_1337 · 2026-10-06
1 near-duplicate retellings: ank_parikh