Summarize this content to 2000 words in 6 paragraphs
Enzyme designers in the David Baker lab at the University of Washington in Seattle, from the left: Sam Pellock, Anna Lauko, Kiera Sumida, David Baker, Donghyo Kim, Indrek Kalvet and Seth Woodbury. (UW Institute for Protein Design Photo / Ian Haydon)
A team from the University of Washington led by Nobel laureate David Baker is using artificial intelligence to design effective enzymes from scratch — an accomplishment the researchers call “one of science’s grand challenges.”
Enzymes are the wizards of the natural world, proteins that can transform molecules and rapidly accelerate chemical reactions under mild conditions. They’re found in every living cell and are essential to life. Enzymes are already being harnessed for drug production and industrial processes. The newly developed tools for enzyme creation could unlock wide-ranging applications.
“Now we can make these enzymes tailored to any reaction of interest, theoretically,” said Anna Lauko, a recent Ph.D. graduate from Baker’s lab. “It’s sort of flipped the way that we would do enzyme design.”
Lauko is a co-lead author of a research paper being published today in the journal Science. Her co-leads are Sam Pellock, an acting instructor in the lab, and Kiera Sumida, one of Baker’s graduate students.
Last year Baker, a biochemist and director of the Institute for Protein Design at UW Medicine, won a Nobel Prize in chemistry for his work unraveling the molecular design of proteins and developing technologies for using AI to build and test new ones.
In the past, scientists made Frankenstein enzymes, stitching together components of existing proteins in the hope that the assembled parts could manage a precise task. But enzymes often need to perform nuanced operations, changing shapes multiple times as they manipulate molecules.
Digital image of an AI-generated serine hydrolase enzyme created by UW researchers in David Baker’s lab. (UW Institute for Protein Design Image)
Pellock compared the old approach to enzyme design to going to a thrift store for a suit.
“It’s unlikely that you’re going to find a suit that fits well,” he said, and the enzymes were the same way. They included the basic pieces, but didn’t perfectly match the molecules that they needed to interact with. The new approach produces bespoke proteins.
To test their cutting-edge approach, the UW researchers focused on a well-studied enzyme called a serine hydrolase. The enzyme is able to cleave a chemical bond that’s key to the structure of many carbon-containing molecules, including plastics, polyesters and a common fat in humans.
The team used the RFdiffusion model, an AI program for generating proteins that was previously developed by Baker’s lab and is open source. They combined that with a newer tool called PLACER that helped them identify the most promising de novo enzyme candidates. The scientists then tested the performance of the machine-created enzymes.
“They’re still not quite as good as native enzymes,” Pellock said. “But out of the computer, these are among the best that have been made and they were made with very high accuracy.”
The accomplishment is a milestone and proves that the researchers are getting close to making new enzymes for human-driven tasks that could out-perform what nature has produced.
Sumida, for example, is working to build an enzyme that could help degrade the planet’s massive glut of plastic waste. Plastic is an incredibly new substance on an evolutionary scale so there hasn’t been much time for enzymes to evolve that can break it down.
There is an enzyme in the serine hydrolase family that can chop up the bonds in the plastic that’s used to make water bottles and other products, but there are many other kinds of plastics out there that need to be disposed of sustainably.
“We thought that it would be a really good application if we’re able to build these enzymes from scratch,” she said, and customize them for different types of plastic.
The researchers are eager for the advent of high-performing, AI-designed enzymes after decades of largely disappointing efforts.
“Hopefully you’ll start hearing more about enzyme design projects,” Pellock said, “because they’ll actually yield a functional enzyme at the end of them.”
Additional authors for the Science paper are David Baker, Ivan Anishchenko, David Juergens, Woody Ahern, Jihun Jeung, Alex Shida, Andrew Hunt, Indrek Kalvet, Christoffer Norn, Ian Humphreys, Cooper Jamieson, Rohith Krishna, Yakov Kipnis, Alex Kang, Evans Brackenbrough, Asim Bera, Banumathi Sankaran and K. N. Houk.
Author affiliations include the following UW departments and programs: biochemistry; biological physics, structure and design; Institute for Protein Design; molecular engineering; chemistry; the Howard Hughes Medical Institute; and the Paul G. Allen School of Computer Science and Engineering. A scientist from UCLA’s Department of Chemistry and Biochemistry also contributed.
