Directed evolution is a method for engineering proteins by mimicking natural selection in the lab instead of designing them rationally from scratch. Researchers only specify the desired end behavior, then run iterative rounds of random mutation on a protein, screening out variants that fail and keeping ones that improve, until the protein evolves the wanted function. Francis Arnold pioneered this approach and won the 2018 Nobel Prize in Chemistry for using it to engineer custom enzymes.
- 01
One host compares directed evolution's iterative mutate-and-select loop to reinforcement learning loops used in AI coding today.
234 words · auto-generated from the episode video
1:28:22Um it abandoned this rigid rational design of proteins where it's like, okay, I want the protein to do this and like I want it to look that way. What Francis Arnold said is, "Why don't we let life figure it out how to make the protein look the way that it wants it to look? I'm only going to optimize for the endpoint behavior and use evolution within the lab to optimize my protein." So, she had these iterative loops that would run random mutations and then she would screen out the mutations that didn't work and filter through the mutations that did. and it it's it's been used to engineer custom enzymes ever since she won the 2018 Nobel Prize
1:29:02in Chemistry. >> Very similar to the concept of Ralph Loops in AI coding. >> Very much the exact same concept. >> We can thank uh directed evolution as part of the inspiration there. 1993 to 1994 discovery of the impact of comet Schumacher Levy 9. >> Yes. So this is astronomer Carolyn Schumacher, Eugene Schumacher, who we had earlier who um figured out that the meteor crater in Arizona was an actual meteor and um David Levy. They discovered comet Shoemaker Levi 9 using the telescope at Palomar Observatory just down south here from Los Angeles. Um this was a comet that was captured by Jupiter's gravity and then torn apart