Enzymes designed with AI
Researchers at the Institute for Protein Design (IPD) at the University of Washington School of Medicine have used AI models to design new proteins that act like molecular scissors. Called proteases, these enzymes cut target proteins at specific points and can repeat the cut over and over.
Anqi Chen, a postdoctoral fellow at the institute, says a molecule can destroy one target and then immediately hop to a second, third and thousands of different targets. That sets it apart from conventional drugs, which bind to a single target, inactivate it and stay there.
Alzheimer's and ALS targets
In one paper, with Chen and Kejia Wu as first authors, more than one-third of the new proteases precisely cut a protein implicated in Alzheimer's disease, doing so 100 million times faster than cuts made without the enzymes.
In the second paper, led by IPD graduate student Hojae Choi, an activated amino acid called cysteine served as the blade. Those proteins had a 19% success rate in cutting their targets and were up to 30 million times faster than uncatalyzed reactions.
Hurdles before these become drugs
The senior author is David Baker, director of IPD and a 2024 Nobel laureate in chemistry. Baker says the research is significant because of how well the enzymes work and the range of potential applications, which beyond medicine include breaking down hazardous pollutants and plastic waste.
Much work remains before these proteins can become drugs for humans. One challenge is ensuring the proteases do not trigger an immune response that attacks the therapeutic before it can help. Another is turning the enzymes on and off to modulate their impact. Choi is studying that immune issue, called immunogenicity, and is working to combine the enzyme's cutting component with a benign-appearing antibody.
