Nik Sharma, BioCorteX - Advanced Simulation Engine for Drug Development | LSI USA '24

Nik Sharma  0:04  
My name is Nik Sharma, I'm CEO and co founder of by cortex. I'm sort of joined by my co founder, Moe. If anybody saw the SpaceX launch approximately two days ago, just as the conference started, lots of the tech that we're actually using to answer the seemingly impossible question of whether this drug is actually going to work for an individual actually comes from that same basis. So ultimately, the question that we're all really concerned about from a pharmaceutical perspective is is this going to work for me now me could be the patient, the individual, the family member, or actually the pharmaceutical company. This is the largest lever in pharma, the annual spend on r&d is, you know, it's it's astronomical. So you know, 244 billion on a global spend. And 139 billion on at the top 20, pharma, but the success rate is only 4%. Now, that's shockingly bad, really, when you look at it. So if you if you look at what happens in aerospace, if you look at an engine program, an engine program is actually very, very similar to designing a drug, it's about it's about the same cost in the region of two to 3 billion, you you have the same, you know, 100 150, very clever people working on the problem. But obviously, we all get on planes, and they very rarely fail. In fact, they always typically succeed. Obviously, there are sort of cases where that hasn't happened. But the way they're able to achieve that is actually by testing in silico. They're able to practice and refine everything computationally and fail in silico, before they go out and actually put that engine onto a plane. Now, this isn't, because the engines are, you know, we understand engines. And so you know, just say that sort of most background is that he was head of AI at Rolls Royce, the aerospace company. So this is something that, you know, I sort of learned as an MD PhD was, was that we don't understand all of all of the physics actually behind engine design. And what we're doing is actually bringing that right first time approach into drug development. And so the aim is to increase the success in sort of clinical trials, by uncovering the drug bacterial interaction. Now, this isn't something that's new. So this isn't, this isn't a new technique. This is something that we're just late to the party on. This is something that we've co evolved with bacteria, every single pharmaceutical that has been approved or anything that anything that any of you take, may have an interaction with, with with the bacteria, now that interaction can occur at different levels. So it sort of can occur in the gut, it can occur, it has interactions with the host physiology, and sort of finally, if in, in tumors themselves, they're actually bacteria that actually can metabolize the drug. And that's something that's been known about for over 100 years. So what our tech does our to our tech is very different. Now, there's a lot of hype about AI, AR ar, ar tech is not a black box, it's a foundational emulator. Now an emulator, if anybody had a misspent childhood and played a lot in computer games, like I did, you know, you can still get the same software for your retiree. And it's a run it on your MacBook, because what it does is that it emulates the hardware. And then you can just run the software on this. And this is something that NASA has been using, you know, for the last 30 years. So what we conceptualize this, this entire framework is as follows. So normally, you know, Drugs aren't working, we know that. If you think about an app, and we all have have iPhones, we've all had the issues with other moving between Android and Apple. If you don't update to the latest operating system, you know, it doesn't work. It's incompatible. What what we're seeing is actually quite simple. All of the assets are actually targeted for for a hardware. They're all designed to work on a human homosapien, which is great. What we're saying is, is that there's there's an element of variance that occurs in between an operating system, which is the bacteria and the geographical and lifestyle approach of that individual. And this is really important, because if you're within, you know, the United States or say in New York, those bacteria will be very different than if you live in Cambridge in the UK, which is where I'm from, or it's spending time, say in India or in in the Middle East. And this is one of the things from a health equality point of view, we do see there is a very large variance in how effective drugs are. But from a pharmaceutical perspective, it's imperative to know because if you're going to run your study in Australia, as many people do, Because of the attractive sort of, sort of framework from a regulatory point of view, and then move that to the US, there's actually quite a difference there. Now, from from, from the tech point of view, the way that we build this up is we have a phenomenally large sort of microbiome based knowledge graphs, about 15 billion connections. So to put that in context, there was a, there was an announcement probably three months ago about the largest sort of biology based knowledge graph in existence at about 5 billion. So we're at least three times larger than that. And you know, it's a rubbish in rubbish out. So we actually, it could be much larger. So that's a really tiny amount of our data, what we do is actually build up these in silico representation. So we build up the hardware building. First of all, in silico cells, we can look at small molecules, or ADCs, we can look at, you know, we can build up an individual so in silico, microbiome or in silico, tumor, and then actually run that phase one study or run that phase two study or the or the multicenter trial. Now, ultimately, this isn't a perfect system, just to say we're not saying that we understand all of the biology, but at the moment, this is occurring in all of the pharmaceutical pipelines to some extent. And at the moment, you know, very, very few people actually seeing this. The difference between this and AI, you know, it's important, you know, NASA work, if you think about SpaceX, and that launch, they did the self landing rocket, on its ninth attempt. Now, from a pharmaceutical perspective, you'd launch 1000 rockets, or from an AI perspective, you'd launch 1000 rockets and get that training data, and then trying to learn which aspects worked, obviously not a feasible situation. So so what what we do, which is very different, which actually maps to right down to the biology is that it's built for that mission critical approach. You don't need training data, which means all we need is the chemical structure. So we're able to track all of large pharma, we're able to, you know, we're our ability to be able to look at different assets is, is, is sort of quite broad, actually, we're very computationally tight. Meaning if you look at the large sort of Gen X Gen AI models, the large language models, foundational models, however you want to label them, you know, people are often burning, you know, there's the, there's the little meme about the company that had a sort of POC for 100k, and sort of blew 500,000 on their charge GBT. Sort of bill for the for the project. It's phenomenal to actually run. So if you look at how we validated this, we've actually validated this in a PK study, this was the published data. This is what our engineers actually determined. The PK was from this drove back to your interaction into approved oncology assets. And this is what the real world measured. PK was in those individuals. With as a, as a clinician, it's like, hey, that's great, but why should I care? You care, because what we then did was actually look at for oncology drugs and looked at the difference between the response in Japan and in the US, what you see in the blue is that there is more drug metabolism in the US population than what there is in the population in Japan. If you look at the one year survival, if you live in the US, your one year survival is 16%, lower than if you lived in Japan. And what we're saying is, is that there's an element of this, which is caused by that drug bacteria interaction, not all of it, but certainly an element of it. We then scaled to this, we can now do you know, we've looked at every single approved oncology drug. And the reason for doing all of them is we understand what degree of drug bacteria interaction can be tolerated by a normal trial design. We're now going through the assets that have failed to see whether they land within that, from that geographical perspective. And from a market perspective, looking at Japan, China, India, and the US, obviously, very large markets. And this is one of our first products is being able to run this in silico. And that's, you know, we've published at all the sort of major oncology and sort of major conferences. Our ability to be able to move at speed is that our average time from hypothesis validation hypothesis generations validation is actually measured in days, and sometimes in hours. We have a phenomenal team. It's a very tight team very different backgrounds from an engineering perspective. Sort of Jackie Hunter is the ex CEO of benevolent AI is our chair with with our pharma experience. We're funded by sofinnova and Hoxton. From a commercial sort of model we you know, we have a lot of things that we can go after we've been very focused on a bespoke, well on a on a fixed fee and then bespoke model looking at assets. We can do this for competition as well. We have a codevelopment model where we can actually take a product all the way through, and we're actually looking at sort of trial rescue as well. We raised a 5 million seed in 20. Tea two we're officially partnered with Bayer we work with emoji in the UK and you know we're part of a number of sort of startup programs we're raising a series a 25 million and thank you