DeepMind's AlphaProtein Novo designs new-to-nature enzymes, but weights stay closed
anshulkundaje · x · 2026-10-06
Google DeepMind shared a preprint on AlphaProtein Novo, a system for generative de-novo enzyme design. Building on AlphaFold, protein language models, and AlphaProteo, the team generated new-to-nature enzymes with state-of-the-art activity on 2 benchmark reactions.
However, researchers note that while the code is on GitHub, the model weights are not open, limiting reproduction and follow-up work.
Related event: DeepMind Unveils AlphaProtein Novo for De Novo Enzyme Design(10 posts)→
More from Research
- Domain-driven development in Lean: using theorem proving to separate what from how — hargup13 · 2026-10-06
- Top 3 frontier AI labs grade CAD benchmarks wrong, scores off by up to 90% — hudzah · 2026-10-06
- SDE Matching extended beyond one-time marginals to learn generative models in physical time — canaesseth · 2026-10-06
- Helmholtz-SDE closes the approximation gap in simulation-free latent SDE inference — canaesseth · 2026-10-06
- Gaussian Flow Dynamics enables simulation-free neural SDEs learning temporal structure — canaesseth · 2026-10-06
- OCBench offers human-like scripted policies for scalable robot BC/RL research — kevin_zakka · 2026-10-06