Rundown 4 — AI + protein engineering (“MultiEvolve”)

59:39Mhm. >> Is the antibbody going to bind tighter? Is the crisper tool going to edit better? These are things that aren't really captured by evolution. But this particular multie evolve paradigm is actually going to do. And what what what's very cool about this is now we've put the lab and the AI in the loop. >> Right. Right. Yeah. >> Okay. like it's one big loop of discovery where the AI is interacting with people in the lab, telling you what to what to do in the lab, the lab is going back to the AI and you can have this iteration happen even faster, right? And the AI framework was able to accurately predict how proteins will

1:00:21function even when several of their amino acids get mutated. Because what you want to do is if you have an amino acid, you want to figure out, oh, if I replace this amino acid with a different one, is it going to make it better or is it going to make it worse? What's more is if I replace these two, is that going to make it better? Because if I just replace this one and it's good and I replace this one and it's good, that doesn't mean if I replace both of them, it's good. There's some synergistic >> landscape. Maybe they're antagonists, right? And so all of these relationships are now something that I can parameterize with this multi-evolve model, right? And bio bioengineers can now develop this new machine learning framework that condenses that problem of

1:01:03protein engineering into a single round of testing. >> And in their test, the model was successfully able to find combinations of mutations that outperformed the original proteins. >> Yep. showing that >> that having it in the loop like that actually has fundamental um improvement value. >> Yeah. Yeah. And it's very cool because the Arc Institute which is in Silicon Valley, the whole point of it is to have like frontier AI capabilities and experimental biologists under the same physical roof. >> That's how they operate and closing that loop between computation and the wet lab. >> Yep. >> This is like their first iteration. It's very exciting to see what they're going to come up with next. This is

1:01:44fascinating because this is something similar that you know my mom works in clinical trials and there there's this similar desire to find ways to incorporate tools like AI into the process to create that feedback that that virtuous cycle and feedback loop but you have to start somewhere. And so it's interesting to hear in a variety of these spaces across both fundamental research and application for example in pharmaceuticals um that people are finding ways to integrate it that is changing outcomes that is influencing >> changing yeah >> influencing outcomes for the positive. I get that everyone wants to say it's

1:02:24wrong because they tried it one time and that means that all of it is wrong but um that's just not how it works. Uh it's not perfect, but it is it is filling in gaps that are significantly meaningful. >> Yeah. >> Great rundown this week. I mean, it's going to be the the first story, the rundown. These are some >> these are cool stuff. >> This is some fantastic stories. We're going to move on to our main story number two. This one is a fundamental physics story. It's about trying to measure one of the smallest things in the universe. We talked about it a lot on this pod, the proton. Uh it's one of the fundamental particles inside the atomic nucleus and we've done it to