AI that designs new enzymes from scratch, both structure and sequence, validated with AlphaFold 3. Lab-tested enzymes produce piperidine and break down DEHP; not yet peer-reviewed.
AlphaProtein Novo at a glance
- What is AlphaProtein Novo?
- AlphaProtein Novo is a model from Google DeepMind. Google DeepMind diffusion pipeline for de novo enzyme design by motif scaffolding, with open code and downloadable weights.
- Who makes AlphaProtein Novo?
- AlphaProtein Novo is made by Google DeepMind.
- Does AlphaProtein Novo work in the EU?
- Available in the EU. Code is Apache-2.0 on GitHub and weights are downloadable from Google Cloud Storage; no EU/EEA exclusion appears in the provided material, though weights terms of use were not reviewed in full. Last verified 14 Nov 2026.
- How much does AlphaProtein Novo cost?
- Open source, open source. Code free under Apache-2.0; weights free to download under separate terms of use. Users supply their own GPU compute. Licence: Apache-2.0 (code); weights under separate AP Novo Parameters Terms of Use.
- Where was AlphaProtein Novo shown?
- In 1 creator video by AI Search. Most recently by AI Search on 11 Oct 2026, at 8:02 in “New AI enzymes, 700+ math breakthroughs, AI for DNA mutations, room temp semiconductors: AI NEWS”.
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
AlphaProtein Novo (AP Novo) is a generative diffusion pipeline for de novo enzyme design via structural motif scaffolding. Its diffusion model co-generates protein structures and sequences, conditioned on a catalytic motif and ligand context.
The repository provides an end-to-end pipeline. It combines the generator with optional sequence redesign using LigandMPNN, structure prediction with AlphaFold 3 (including a leaving-atom fine-tuned variant for covalent intermediates), and evaluation metrics. A single run_pipeline.py script runs all stages from a manifest file, and example campaigns such as a Kemp eliminase benchmark are included. A bioRxiv preprint describes the method; creator coverage reports lab-tested enzymes (e.g. for piperidine production and DEHP breakdown) that are not yet peer-reviewed.
Code is released under Apache-2.0. Pretrained weights are not in the repository and are downloaded from Google Cloud Storage under separate terms of use, as are the AlphaFold 3 weights. It requires a Python 3.12 environment with JAX (GPU recommended), and is aimed at computational protein and enzyme design researchers.
- Diffusion model co-generating protein structure and sequence
- Conditioning on catalytic motifs and ligand context
- Optional sequence redesign with LigandMPNN
- Folding and validation with AlphaFold 3 (leaving-atom variant)
- End-to-end run_pipeline.py driven by a manifest file
- Built-in evaluation metrics and example campaigns
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Similar AI tools
Closest neighbours by meaning, not just by category.
AlphaFold 3
Model · Google DeepMind
Google DeepMind and Isomorphic Labs' model predicting structures of proteins, nucleic acids, ligands and their complexes.
LOGOS
Model · Alibaba Tongyi Lab
Autoregressive scientific model that tokenizes proteins, molecules, materials, antibodies and reactions into one shared grammar.
i1
Model · Princeton
Fully open 3B text-to-image diffusion model from Princeton with released weights, training code, and data pipeline.
Claude gene discovery (novel enzyme system)
Research · Anthropic
Anthropic research result: Claude agents found a CRISPR-like array-associated reverse transcriptase system in phage DNA.
AlphaGenome Atlas
Research · Google DeepMind
Searchable database of AlphaGenome predicted effects for every possible single-nucleotide change in the human genome.
NitroFusion 2
Model
One- and two-step image generation model distilled from Z-Image-Turbo using text-stream adversarial distillation.
