RFdiffusion3 designs active phosphotriesterases and transthyretin-specific zinc proteases zero-shot
RFdiffusion3, a generative model that diffuses free atoms to satisfy atomic constraints, produced de novo phosphotriesterases with a two-metal active site and zinc proteases specific for transthyretin, obtaining active catalysts of both kinds from zero-shot design campaigns of 96 constructs.
- Model
- RFdiffusion3
- Field
- Chemistry
- Date
- 2026-10-15
- Human collaborators
- Jasper Butcher, Rohith Krishna, David Baker
Sources
Original work
What was found
The method places individual atoms rather than residues, aiming to meet sets of atomic constraints such as an enzyme active site more quickly and accurately than earlier methods, across tasks from enzyme design to RNA structure design. The experimental demonstrations are two enzyme classes: phosphotriesterases whose active site includes a carbamylated lysine coordinating two metal ions, and zinc proteases that cut transthyretin, the protein involved in transthyretin amyloid disease. In both, a campaign of 96 constructs tested without iterative optimisation yielded active catalysts.
Novelty check
Europe PMC, 2026-10-29: TITLE_ABS:"RFdiffusion" AND TITLE_ABS:"enzyme" returned 17 records, including competing generative enzyme-design methods. Within this registry 2025-02-13-designed-serine-hydrolases records the same institute's earlier diffusion-designed serine hydrolases; this result is a new model and different enzyme classes, a binuclear metal phosphotriesterase and a substrate-specific protease.
Caveats and known objections
Peer-reviewed in Science, with the experiments done by the developers at the Institute for Protein Design and collaborators at institutions including Cambridge, ETH Zurich and the Technical University of Denmark; no independent replication. The abstract gives no catalytic efficiencies, hit rates or structural validation, so how active the catalysts are relative to natural enzymes is not stated there. The protease is a laboratory result with no clinical endpoint. Autonomy is search-scaffold, as for the earlier designed hydrolases: the reactions, active-site geometries and assays were human-specified and the model generated designs inside that pipeline.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and Search scaffold.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and search scaffold for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). RFdiffusion3 designs active phosphotriesterases and transthyretin-specific zinc proteases zero-shot. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-15-rfdiffusion3-enzyme-designBibTeX
@misc{whataifound-instituteforproteindesign-2026-design,
title = {RFdiffusion3 designs active phosphotriesterases and transthyretin-specific zinc proteases zero-shot},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Institute for Protein Design, University of Washington. Verification: Peer reviewed. Autonomy: Search scaffold.},
url = {https://whataifound.org/finding/2026-10-15-rfdiffusion3-enzyme-design}
}