The Pulse
UW Teams Design Enzymes to Cut Disease-Linked Proteins
University of Washington researchers reported two AI-designed protease systems in Nature on October 14, 2026. One cut selected protein targets in mammalian cells; the other targeted disease-associated proteins including TDP-43 and amyloid-b
AI.info Team · Written with AI ·
Two enzyme designs from University of Washington researchers can cut selected protein bonds, but their strongest results come from different settings: one system cleaved targets in mammalian cells, while another was designed to attack proteins associated with disease. The studies, published in Nature on October 14, 2026, show progress on a long-standing protein-design problem without demonstrating a treatment in patients.
Two chemistries tackle the same cutting problem
The teams designed proteases, enzymes that break peptide bonds in proteins, using computational methods to specify where a cut should occur. One approach built an enzyme around an activated cysteine amino acid. The other used a zinc ion to help activate a water molecule that attacks the bond.
Unlike a drug that binds to a target and blocks it, an enzyme can cut a target and then act again. That makes designed proteases a possible route to changing or removing proteins rather than simply holding them in place. The experiments, however, test distinct steps toward that prospect; the results should not be read as showing that every disease target was cut inside cells.
“Designing proteases from scratch has been a longstanding challenge. It is really exciting that we created entirely new cysteine proteases that cleave the chosen bonds specifically, and they also work inside living cells. This is a step toward designing tailor-made molecular scissors,” said Hojae Choi, a Ph.D. student in biochemistry at the University of Washington School of Medicine and a lead author of one of the studies.
RFdiffusion produced active zinc enzymes
In “Computational design of metalloproteases,” researchers used RFdiffusion to build zinc-based enzymes around a catalytic site and a binding pocket shaped for a chosen peptide sequence. Of 135 designs tested, 36% catalyzed cuts at their intended bond. The best design accelerated peptide-bond hydrolysis by more than 100 million times compared with the uncatalyzed reaction; four additional substitutions raised that acceleration above 10 billion times.
The researchers designed enzymes to cut TDP-43, amyloid-beta and serum amyloid A, proteins associated with human disease. Rate enhancements for those targets reached as high as 1 billion times, according to the paper. The study also used the proteases to activate engineered, caged cytokines and receptor antagonists, allowing researchers to control cell state through a designed molecular trigger.
Those results do not mean the enzymes are ready to treat Alzheimer’s disease, ALS or other conditions. The paper says the designed proteases’ turnover rates remain about 1,000 to 10,000 times below those of the most efficient natural proteases under similar conditions. The authors also identify catalytic efficiency and selectivity as areas for further improvement.
Cysteine proteases cleaved targets in mammalian cells
The second report, “De novo design of cysteine proteases,” tested enzymes built around an activated cysteine that serves as the catalytic nucleophile. Thirteen of 69 designs worked, a success rate of 19%, and the best accelerated the background reaction by up to 30 million times. Their folds differ substantially from those of naturally occurring proteases.
The study reports that the designed enzymes cleaved their targets in mammalian cells. UW Medicine’s account says researchers also tested sequences drawn from proteins implicated in Alzheimer’s disease and amyotrophic lateral sclerosis. That distinction matters: the published findings establish designed cleavage and cell-based activity, not a clinical effect or a demonstrated therapy for either disease.
Cell experiments are not yet treatment results
The two papers offer separate pieces of evidence: one reports disease-protein targeting and protease-triggered control of cell state, while the other reports protease activity inside mammalian cells. Taken together, they show that computational design can produce enzymes that cut chosen sequences rather than merely bind them. They do not establish how the enzymes would behave as medicines, including how they could reach the right tissue or avoid unwanted cuts.
For now, the concrete milestones are laboratory results: 13 active cysteine-protease designs out of 69 tested, and 36% activity among 135 zinc-protease designs. The disease-associated targets and the cell experiments mark different accomplishments across the two studies; neither paper reports a clinical trial.