AlphaProtein Novo Designs Enzymes From Scratch
Google DeepMind’s AlphaProtein Novo is a generative diffusion pipeline that designs enzymes around catalytic motifs and ligand contexts instead of modifying natural proteins. Its preprint reports new-to-nature piperidine synthesis, DEHP degradation, and strong results on benchmark reactions.
AlphaProtein Novo makes de novo enzyme design look less like a moonshot and more like an emerging engineering workflow, though experimental validation remains the bottleneck.
- –The pipeline combines diffusion-based structure and sequence generation with LigandMPNN, AlphaFold 3, and mechanism-aware filtering.
- –Its lead piperidine synthase reportedly achieved 99:1 regioselectivity, showing AI-designed proteins can control reaction outcomes nature does not naturally favor.
- –DEHP degradation at conditions that denature natural enzymes points toward practical environmental and industrial applications.
- –The released code improves reproducibility, but model weights require separate access and are subject to usage restrictions.
- –The work still depends on laboratory screening and directed evolution, making AI a powerful starting-point generator rather than a fully autonomous biochemistry lab.
DISCOVERED
1h ago
2026-10-06
PUBLISHED
1h ago
2026-10-06
RELEVANCE
AUTHOR
codvynn