TL;DR: On Oct. 14, 2026, University of Washington researchers reported two sets of computationally designed proteases in Nature that cut chosen protein sequences with high speed and precision.

Researchers at the UW Medicine Institute for Protein Design built synthetic enzymes that act like programmable scissors for proteins. The team selected target sequences first, then created proteins intended to cut those sequences, then checked cutting accuracy in laboratory tests. According to the UW Medicine Newsroom article titled “Molecular ‘scissors’ open path to new therapies,” the work appeared in two studies in the journal Nature on Oct. 14, 2026.

The two Nature papers describe different chemical strategies for the same broad task. One paper, titled “De novo design of cysteine proteases” and led by Hojae Choi, describes enzymes centered on an activated cysteine. The other paper, titled “Computational design of metalloproteases” and led by Anqi Chen and Kejia Wu, describes enzymes that use a bound metal ion for cutting. Both approaches used computer-based design followed by experimental screening.

Two enzyme types reported on the same day

The cysteine-protease study tested 69 designs and found activity in 13, a result reported as a 19% success rate. The metalloprotease study tested 135 designs and found that 36% catalyzed cleavage of their intended peptide substrates at the intended bond. The UW Medicine Newsroom summary reports the same two success figures and identifies the lead authors for each study.

The reported speed increases were large when compared with reactions without a catalyst. The cysteine-protease paper reports rate enhancements, expressed as kcat/kuncat, of up to 3 × 10⁷. The UW Medicine Newsroom summary describes the same result as cutting a target sequence up to 30 million times faster than the uncatalyzed reaction.

For the metal-based designs, the Nature abstract reports that the best design accelerated peptide-bond hydrolysis by more than 10⁸-fold relative to the uncatalyzed reaction. It further reports that four substitutions increased the acceleration to more than 10¹⁰-fold through improved active-site preorganization. The UW Medicine Newsroom summary highlights a related result for amyloid-beta, stating precise cutting at 100 million times faster than the uncatalyzed reaction.

What the enzymes were shown to cut

The researchers tested the new enzymes on purified proteins. They also tested sequences taken from proteins linked to Alzheimer’s disease and amyotrophic lateral sclerosis. In one of the studies, the team also tested function inside living human cells grown in the laboratory. The UW Medicine Newsroom article lists those three experimental settings together.

The metalloprotease Nature abstract names three human disease-associated proteins that designed enzymes were made to cut selectively: TDP-43, amyloid-beta and serum amyloid A. It reports rate enhancements of up to 10⁹-fold for those disease-related targets. The UW Medicine Newsroom summary specifically identifies amyloid-beta as a protein implicated in Alzheimer’s disease.

The cysteine-protease Nature abstract states that the designed proteases efficiently cleaved their targets in mammalian cells. That finding is presented as opening the door to synthetic-biology applications. The UW Medicine Newsroom account likewise says the cysteine proteases worked inside living cells.

How design differed from natural proteases

The cysteine-protease designs did not copy the shapes of known cutting enzymes. The Nature abstract reports that the designed proteases have folds very different from proteases in nature, with TM score below 0.50. It also reports that six crystal structures were very close to the design models, with C-alpha root mean squared deviation below 1.2 angstroms.

Those structural results were presented as evidence for both broad design ability and model accuracy. The Nature abstract links the unusual folds to capacity for generalization and links the close crystal-structure agreement to accuracy of the design method. Experimental and computational analyses pointed to remaining differences from the most active natural cysteine proteases, attributed to imperfections in active-site preorganization and substrate positioning.

The metalloprotease Nature abstract says the team used RFdiffusion methods to design proteins with water-activating zinc-based active sites next to a peptide-substrate binding pocket. The UW Medicine Newsroom summary describes the same idea in simpler terms, saying one group built enzymes around an activated cysteine blade while the other used a bound metal ion rather than an amino acid as the cutting blade.

Why precise protein cutting is difficult

Protein-cutting chemistry requires very exact placement of the active components. The UW Medicine Newsroom article says the blades of protein scissors need near-atomic precision to enable cutting chemistry. It also says natural proteases use subtle flexibility to grip their target sequence. It identifies both requirements as a significant design challenge.

The article attributes recent progress to newer computational models. It says those models improved the ability of protein designers to handle the placement and flexibility problems. The researchers used deep-learning-based design methods to develop structures, screened the candidates for cutting of a specific target sequence, and confirmed that the resulting protease structures matched their computational blueprints.

David Baker, director of the institute, professor of biochemistry at the University of Washington School of Medicine and Howard Hughes Medical Institute investigator, said the field is moving from predicting and building protein shapes and interactions toward designing machinery that changes matter at atomic level in targeted and programmable ways. The UW Medicine Newsroom article quotes him calling the result “a major step forward for protein design.”

Enzymes versus binding medicines

Most medicines act by binding to a target. The UW Medicine Newsroom article explains that a small molecule or antibody latches onto its target and holds on, blocking or activating it one molecule at a time. It says that one-to-one relationship sets a medication limit because engaging one million copies of a disease target generally requires one million or more copies of the drug.

That arithmetic helps explain dosing and delivery for some protein medicines. The article says many protein medicines must be given in substantial quantities by infusion, an experience familiar to patients who spend hours receiving intravenous fluids or who live with an implanted vein-access port. The comparison is used to explain why a reusable enzyme could behave differently.

An enzyme changes its target and is released unchanged and ready to act again. The UW Medicine Newsroom article says a single designed protease can cut thousands to millions of specified targets over its lifetime. It says that reusability allows the same effect at a substantially lower dose than a medicine that must remain bound to each target.

Hojae Choi, a doctoral student in biochemistry at the University of Washington School of Medicine and a lead author of one study, said the team created entirely new cysteine proteases that cleaved the selected bonds specifically and also worked inside living cells. The UW Medicine Newsroom article quotes Choi describing the advance as movement toward tailor-made molecular scissors. Choi’s paper is the cysteine-protease Nature study reporting the 13-of-69 result.

Targets once called undruggable

The advance may matter for proteins that lack a convenient binding site. The UW Medicine Newsroom article says the work opens the door to targeting disease-related proteins traditionally considered undruggable because they offer no pocket for a conventional drug to latch onto. A designed protease does not need the same kind of pocket because it is built to recognize a sequence and cut it.

Anqi Chen, a postdoctoral scholar at the University of Washington School of Medicine and co-lead author of one paper, said proteases are only the beginning. The UW Medicine Newsroom article quotes Chen saying the team envisions designed molecules that act on biological systems in increasingly sophisticated ways, creating new possibilities for studying biology and treating disease.

The possible therapeutic uses listed in the UW Medicine Newsroom article remain examples rather than demonstrated treatments. They include destroying a viral protein or bacterial toxin outright, clearing a misfolded protein that accumulates in neurodegenerative disease, and snipping a protective cap from a drug so it functions only in needed tissue. The article presents those outcomes as approaches that binding agents could not achieve and that functional proteases could enable.

Bioorthogonal control and caged proteins

The metalloprotease study went beyond direct cutting of disease proteins. Its Nature abstract reports that the researchers designed caged cytokines and receptor antagonists that are selectively activated by the designed proteases. That system is described as enabling bioorthogonal control of cell state.

In that context, bioorthogonal means the designed protease and its intended protein switch operate alongside natural biology without relying on natural cutting signals. The UW Medicine Newsroom summary does not add dosing, safety or human-test details for the caged-protein system. The Nature abstract presents the caged cytokines and receptor antagonists as part of the same programmable-protease platform, alongside applications in therapeutics, biotechnology and bioremediation.

The cysteine-protease Nature abstract separately points to synthetic-biology applications after showing efficient cleavage in mammalian cells. Neither Nature abstract, as summarized in the available reports, claims human treatment, clinical efficacy or an approved medicine. The demonstrated settings are purified proteins, disease-linked sequences, cultured human cells and mammalian cells, depending on the study.

What was verified and what was not

The verified core is design, cutting precision and speed in controlled tests. Researchers specified sequences, built candidate enzymes by computation, screened for cutting, measured acceleration over background, checked structures against models and tested selected enzymes on disease-linked sequences or in cells. The two success rates, 19% for the tested cysteine designs and 36% for the tested metalloprotease designs, describe laboratory screening outcomes, not patient outcomes.

The structural verification was strongest for the cysteine group. Six crystal structures matched design models within 1.2 angstroms C-alpha root mean squared deviation, while the overall folds remained dissimilar from natural proteases by the reported TM-score threshold. The metalloprotease paper emphasized functional verification across 135 tested designs and further improvement through four substitutions in its best design.

No human dosing, side-effect profile or clinical benefit was established in the materials summarized here. The UW Medicine Newsroom article describes lower-dose potential as a consequence of enzyme reusability, not as a measured clinical dose. It describes infused protein medicines and vein-access ports only to explain the burden associated with one-to-one binding medicines.

Funding and research teams

The work came from the UW Medicine Institute for Protein Design. The cysteine-protease author list includes Hojae Choi, Magnus S. Bauer, Brian Coventry, Preetham Venkatesh, Anqi Chen, Donghyo Kim, Samuel J. Pellock and David Baker, among others. The metalloprotease author list includes Anqi Chen, Kejia Wu, Hojae Choi, Preetham Venkatesh, Samuel J. Pellock, Donald Hilvert and David Baker, among others.

The UW Medicine Newsroom article says the work was supported by the Howard Hughes Medical Institute, Audacious Project, Open Philanthropy Project Improving Protein Design Fund, Schmidt Family Foundation, Wu Tsai Translational Investigator Fund and others, with full funder lists in the published papers. The Nature records list data and code availability, including Protein Data Bank deposits and public GitHub and Zenodo entries, but the UW Medicine Newsroom summary is the source used here for the biological interpretation and therapeutic context.

Together, the two Oct. 14 Nature studies expand computer-designed proteins from binders and shapes to active, reusable cutting tools. The next questions concern durability, selectivity in complex tissues, delivery and safety, none of which are answered by the cutting and cell-culture results reported so far. For now, the documented advance is precise, programmable cleavage: selected sequences cut at selected bonds, verified in test tubes and cultured cells, at speeds millions to billions of times above background.

Sources referenced