How Physical AI Is Changing Drug Discovery | Jack Hidary on CNBC

Very exciting about this NVIDIA partnership. It's being announced today. We're breaking news here on Squawk on the Street at BIO. BIO is the 20,000-person conference in San Diego, focused on the biopharma industry. We're really excited about this partnership because it allows drug companies, biotech companies, and medical centers and researchers to use a combination of BioNeMo, which is a really strong platform from NVIDIA for the biopharma market, and AQState. AQState is a large quantitative model from SandboxAQ, and it really helps drug researchers, drug discoverers, figure out in particular whether a drug will work against the GPCR. That's the protein that you go for in Ozempic, in Mounjaro. In fact, Sarah, more than 30% of all new drugs that are going in front of the FDA are focused on this target in the body, GPCR, and that's where we're focused in this partnership of NVIDIA and SandboxAQ.

It took literally over a decade, decades in some cases, to hit this target before, and that led to Ozempic, that led to Mounjaro. Now, there's a whole new class of drugs that want to hit this target. And with this virtual screening tool, instead of taking years to decades in the laboratory, you can do this in weeks and months to screen for the kind of molecules known as ligands that will actually bind to the GPCR protein. What's new about this is not only can we figure out whether it will bind and how it will bind, but also whether it's going to turn on the protein or turn that protein off. In the case of Ozempic, you want to turn it off so that you control hunger in the case of diabetes as well. In other cases, you may actually want to turn it on. And so those are agonists and antagonists. And this is a breakthrough between NVIDIA and their BioNeMo platform and our AQState, one of our quantitative models. As you know, we really have to go beyond language models if we want to tackle the biggest problems in medicine.

Yeah, biotech companies. Quite a number of companies will be at BIO, and NVIDIA will be announcing some of the companies that are already starting to use this combination. Drug companies, researchers in the academic setting will also be using it. It's downloadable. People can start to use it. The combination of AQState from SandboxAQ and BioNeMo. And the reason why this is so novel is that these are physics-based models. Jensen has often talked about AI for the real world, physical AI. This is part of that ball game where we're not talking about training on language, we're talking about training on the actual science and physics of our world. That's what delivers this new breakthrough announced today.

Carl, that's exactly correct. The reason why it's been so difficult in drug discovery is that it takes so long to synthesize each potential new candidate drug, test it in a laboratory. Now we can move to virtual screening and testing, but with models that are really high fidelity with our bodies, with the receptors in our actual human physiology. This is really a fundamental new way of discovering drugs, not just cutting down on time, but Carl, to your point, moving away from drugs that will not work, that won't have the binding affinity, that won't have the right mechanism of action, to ones that do. Really excited to announce this today with NVIDIA.