Rundown 3 — prenatal hormones, 2D:4D ratio, and evolution

54:36surprising driving factor to brain size and finger size. The clue lies in 2D to 4D finger ratio. Okay, that's the ratio between your ring finger and your index finger. Okay? And if you've got a longer index finger, that usually means that you've had higher prenatal estrogen exposure during the first trimester. And that's also related to larger brain size. >> Boo. >> Um, >> for boys and not girls, too. >> Oh, no. I have a small brain. >> Now, I think I think this is an overarching evolution story rather than [laughter] individual. Fair, very fair,

55:16very fair, very fair. >> You know, because because I think I think what they're saying is the effect is something where these two genes are correlated. >> Okay. >> Right. >> Okay. >> And it's in line with a lot of um recent research that shows that there's like this this thing called the feminization of the skeleton. And this is not like anything political or anything like that. >> It's not a culture war, I think. >> No, this is not a cultural war at all. It's relevant because in human evolution, increases in brain size are found alongside something called the feminization of the skeleton. For example, um high values of this 2D to 4D in males, the the the ring finger

55:59to the um index finger, they've been found with elevated rates of heart problems, poor sperm counts, predisposition [clears throat] to schizophrenia in males. I see. >> Okay. And it seems that this stuff is coming from prenatal hormones from the fact that we are >> um we were subjected to estrogen in our first trimester that led to heightened brain size but it also led to all of these problems. >> And in the course of evolution, >> I guess the human species decided I'll go I'll go with the large brain size even though it comes with all of these problems. >> That's that's fascinating. No, that that that isn't the I I you know the I'll let

56:41me not I'll leave it. Yeah, I I won't let me leave it. >> This is out of the Swansia University and Istanbul University in the journal Early Human Development. I thought it was I thought it was quite interesting because you know we always think about how genes are related to each other when they when they go through evolution. So So this is something that's very important. I mean, you could you can also map this on to this ide, you know, when people talk about quote unquote masculinity and the change in body, the body stature of like men generally over time, there's this if there's a correlation higher brain size equals to the skeletal structure of us being slightly different. That is an interesting >> that is an interesting angle. Yeah. >> But that like we kind of see manifesting

57:21unfortunately in the culture war, but has sort of this uh genetic uh basis for it. Fascinating story number three. Our last story of the rundown combines AI with bioengineering. So, no no risk here at all. A new machine learning pipeline that is streamlining protein engineering. We've talked about AlphaFold uh God knows how many times. Yes, >> we've gotten comments. Yes, we understand it's not perfect. Nothing we talk about on the podcast is perfect. >> No, this is science. >> We get that. Um, that being said, >> that being said, alpha fold is not everything, >> right? >> Right. Alpha fold means you give me a

58:02protein sequence, which is a bunch of amino acids in a line, I'll probably be able to tell you what it looks like in 3D. >> Okay. And how it changes shape. >> Mhm. >> Okay. What if we wanted to design proteins? If we want to design proteins, the search space is incredibly big. Because if I have a 100 amino acids that I want to fit into a protein, there's 20 choices for each amino acid. That's 20 to the 100th power, there's more atoms in the universe. So, I'm not going to like I I need ways to like hone down if I want to design a new protein. >> Mhm. >> Right. How to actually reduce that search space. >> Right. >> Okay. >> Right. So nowadays what we can do with

58:45various techniques is get down to like tens of thousands of protein candidates. If we get down to tens of thousands of protein candidates, the bottleneck then becomes the lab, >> right? >> Because I can only really efficiently build and test around a hundred different variants, >> right, >> of my proteins. So what's the best way to choose which hundred I want to actually test and make in the lab and then figure out? That is what these guys at the University of California at Berkeley and the Ark Institute are doing with this particular story that came out in science. They created something called multi-evolve. >> The idea is they they've created an ensemble of protein language models.

59:26>> And what these protein language models do is not just care about the fitness landscape of like what is the stuff that evolution constrains me for. >> These models also care about stuff like is the enzyme going to catalyze faster?