AI helps scientists design enzymes that cut chosen proteins
Scientists used AI tools to design new enzymes that snip chosen proteins, Ars Technica and GeekWire reported. The work came from David Baker's lab at the University of Washington. Some of the enzymes cut proteins tied to brain diseases, but any drug use is still far off.
Researchers at the University of Washington used AI tools to design enzymes that cut specific proteins, Ars Technica and GeekWire reported. The work is described in two papers published Oct. 14 in the journal Nature.
The enzymes are called proteases. They break the bonds that hold proteins together. Without an enzyme, a typical bond of this kind would last for hundreds of years in water, Ars Technica reported.
The team is led by David Baker, director of the Institute for Protein Design and a 2024 Nobel laureate in chemistry, GeekWire reported. Anqi Chen and Kejia Wu are the first authors of one paper, and Hojae Choi is the first author of the other.
The scientists started with the cutting core of two natural enzymes, one from a crayfish and one from yeast, Ars Technica reported. Software called RFdiffusion2 built a frame around that core, and another tool, ProteinMPNN, filled in the building blocks.
About 19% of the designs in one project worked, both Ars Technica and GeekWire reported. In the other project, 35 of 38 designs showed some cutting activity once the team placed a water molecule in the right spot, Ars Technica reported.
The team also designed enzymes aimed at proteins tied to Alzheimer's disease and frontotemporal dementia, Ars Technica reported. One design cut its target inside cells.
"You have this catalytic capability to destroy one target, and then immediately your molecule can hop over to the second, third and thousands of different targets," Chen said, GeekWire reported. That could let a drug work at very low doses, she said.
The researchers are now working on targets for malaria, chronic diseases and cancer, GeekWire reported. Baker said the enzymes might one day also break down pollutants and plastic waste.
Big hurdles remain. A patient's immune system could attack the enzyme before it helps, GeekWire reported. Choi is working on a fix that pairs the cutting part with an antibody.
Proteins are also hard to deliver to where they are needed, Ars Technica reported. It added that the AI tools are good at building shapes but cannot yet be asked for a specific catalytic activity.
What we know: UW researchers used AI tools to design working proteases, described in two Nature papers on Oct. 14.
What we don't: Whether the enzymes can be made safe enough, and delivered well enough, to become medicines.
Where outlets agree Ars Technica and GeekWire both report that about 19% of the designs in one of the two projects worked.
Where they differ The outlets cited here report the same key facts so far.