Origin is using AI to accelerate drug discovery by predicting molecular behavior

before we're we're at that level. Congratulations. We'll talk to you soon. Have a great rest of your day. Great uh great to catch up, Chuck. Uh quickly, let me tell you about Finn.

ai, the number one AI agent for customer service, number one in performance benchmarks, number one in competitive bakeoffs, number one ranking on G2. We have our next guest. We've been keeping him waiting. We have Yash from Origin. Uh introducing Axis. Let's bring him in from the reream waiting room.

Yosh, how you doing? Hey guys, thanks for having me. Am I pronouncing that correctly? It's Yosh. Yeah, Yosh. Fine. Uh, introduce yourself. Introduce the company. Give me the news. Yeah. Um, so I'm the co-founder and CEO at Origin. Uh, we're a new startup based in San Francisco.

And Origin is developing uh AI systems to um develop drugs for complex diseases. And today is exciting because we announced the release of Access um our first model. Amazing. Uh give me the performance metrics. What was the benchmark and how'd you do? Yeah.

Um so Access is outperforming Google Deep Minds Alpha Gino on Wow. Jord, did you see what Google stock did today? $20 billion erased from their market cap. It's down half a percent. And I think it's because you did. Look what you did. Look what you did. Congratulations.

Jokes aside, uh, it is impressive to outperform Deep Mind at anything, let alone something as complicated as this. Uh, how did you do it? Is it a function of of scale, a new algorithm, some fundamental insight? Are you doing tech transfer from university? Like h like what is the origin of origin?

Yeah, I mean I I think like large credit goes to the team. um because the team is you know composed of computer scientists, math majors, biologists and it's these ideas coming from um from various fields. Um in terms of the model um the idea was simple.

We wanted to unify a lot of biological modalities uh and a lot of capabilities into one single base model. Most of the biomodels out there um they're extremely marginalized. perform one specific task.

But biology is this one domain that sort of warrants um you know this unified capability because uh if you look at most cells um it's basically a lot of information flowing within cells between cells and you have all of these moving parts and it's an extremely complex system.

Um so out of all the fields it's the one that that requires unification of all these capabilities and our model is the first to do that.

Talk to me about uh when I think about like technology in bio, I think about this spectrum from uh Alphafold, which was Nobel Prize winning, but ultimately didn't really move the biotech markets. I think it was eventually open- sourced.

It hasn't become this like powerhouse enterprise software company that's worth billions and thrown off free cash flow. And then you have a company like Benchling, uh an electronic lab notebook. It's SAS for biotech companies.

It is in the cash flow machine probably I don't know but they're they're making revenue they're charging people they are directly interfacing with biotech companies as customers making revenue how do you see yourself now are is this more of a foundation model lab company you're doing research and then you hope to commercialize it create a product around it or maybe there will be an entirely new novel idea that comes out of this like how chat GPT came out of a bunch of LLM research that was kind of looking hopeless for years and then all of a sudden was the most valuable thing How are you thinking about where you are on that curve between like science, open- source research papers and just SAS?

Yeah. Um, so we trained origin or we trained access as a first step to optimizing the design of gene therapies. Want to make these therapies safer. We want them to have this increased efficacy.

So our focus now is going to be on expanding the models capabilities um to encompass the various sort of components that go into designing these therapies.

uh and also taking the model into the vet lab to actually study the sequences the model is designing uh and it is completely our intention to have a therapeutic program within one year uh where we're targeting diseases already.

Um so the focus is to sort of close this loop train the best models in the world and get therapies out to patients. You're going to do you're going to do it in self yourself agree that's exciting. That's what I was going to ask.

We we uh you know feel do you do you expect uh how do you expect the FDA to have to evolve to um new capabilities on the sort of simulation side because we've talked to a number of you know we've talked to to founders that are developing drugs on the show and they say like you can simulate you know whatever you want but eventually you have to test and then you have to test it in dog or monkey dog monkey mice eventually get it into human and uh there's quite a lot of time in order to really drive those feedback loops.

So, do you think the FDA will will try to um will have to evolve at all or can you work within the current system? Yeah, I I think there's already like positive indications of this. Um the FDA, they want to move away from um animal talk studies for monoconal antibodies. Uh so that makes a good first step.

But in order to sort of really make this happen, we have to make these deep learning systems better because you want to be able to sort of recapulate everything that's going on within these biological systems within tissues uh and then eventually within entire organisms.

Um so I think it's going to move along with the technology. Um so as the technology gets better, we probably expect um you know uh new policies, new regulation coming out. Well, congratulations on the progress.

come back on anytime you have news and if you ever if you're ever develop anything for uh a drug for amateur bodybuilders John would love to join uh if you can't find any monkeys to test on. I'm happy to be a guinea pig. Yeah, send it over. Uh thank you so much for coming on the show. We'll talk to you soon. Awesome.

Yeah, thanks for having me guys. See you tomorrow.