The research team, helmed by David Baker—the recipient of the 2024 Nobel Prize in Chemistry—has successfully employed AI tools to design two distinct types of catalytically active proteases from the ground up: cysteine proteases and metalloproteases. The design of cysteine proteases achieved a success rate of roughly 19%, with the pinnacle of catalytic acceleration reaching approximately 30 million times. In the case of metalloprotease designs, the success rate soared to about 36%, with the highest catalytic acceleration surpassing an astonishing 10 billion times. The actual structures of these designed proteins closely matched the predictions, showcasing catalytic functionality without the need for natural enzyme scaffolds as a crutch.

Validation processes confirmed that these artificially crafted proteases were capable of executing predefined protein cleavage tasks within mammalian cells. Moreover, they demonstrated the ability to selectively cleave proteins linked to a variety of diseases, including TDP-43 and β-amyloid. Beyond this, these proteases could also serve as molecular switches, facilitating bioorthogonal regulation of biologically active molecules such as cytokines and receptor antagonists. This groundbreaking achievement surmounts the technical hurdles that have long impeded the de novo design of enzymes with specific catalytic activities. It paves the way for promising future applications in the development of conditionally activated therapeutic proteins, as well as in the realms of biotechnology and bioremediation.