Video Transcription
Ali Tinazli 00:03
So I'll be speaking today about personalized medicine and how to build a generative AI and then ultimately, to use information for health. You all know the physician's best friend. It's a stethoscope invented 200 years ago. Actually funny to watch that physicians are still using it. Why do physicians use it? Because it's a disease-agnostic, hypothesis-free device. It's actually great. Now think of a technology that is similar to the stethoscope but goes much deeper into health, which basically enables looking into prevention, looking at markers in the context of preventative care, looking into diagnosis at an early stage, also to monitor staging of diseases, and ultimately, also to monitor treatment. Everyone wants to know if the medication I'm getting, the treatment I'm getting, is really the best for my disease. So this all needs to be addressed in a much more personalized fashion, and I will present today basically how we are doing it, by integrating nuclear magnetic spectroscopy with AI and basically to make this an affordable, highly accessible information layer for life spin use in healthcare. Now we speak about life spin use. Life spin use means also lots of dollars, right? So we are looking here in total, basically at a market of about $400 billion we can address with this technology. We are focusing right now, initially on the Asia Pacific market for many different reasons. And what's exciting is that, in terms of the business model as a software as a service, along with the attractive gross margins that you get with the software service business. Now, how do we do it? Right? So what's the magic? So we all know about the genome, so the genome tells you something about the potential of a certain outbreak of a condition in the future or in the near future, but the genome cannot tell you for sure if that will really happen, because environmental factors are another aspect. Then you look at just like in pharmacogenomics, for example, you look into the transcriptome, you look at genetic activity, which is also a very informative layer. But again, it tells you only what should happen. When you look at the level of the proteins, the proteome, then you get a better idea about what's really happening actually in the world, but still, ultimately, where all the different omics are condensing, the end point is basically the metabolites. And what we do is basically we capture quantitatively hundreds of metabolites in a couple of minutes. I will show you quickly how this exactly works, because we believe that the metabolome, so the sum of all the biochemical reactions, represents the actual phenotype of a patient, for example. Now this metabolome is a real-world or it's a real-time phenotype information layer. This information may contain all the important aspects to understand preventative care, diagnosis of diseases, and monitoring of treatment. Now, what have we done so far? So we have now digitized hundreds of thousands of humans. So we have a cyber-physical approach. I will show quickly how this looks like in the movie. So we have a cyber-physical approach where we turn blood into a digital blood that is possible with advanced detection like NMR in a magnetic field. It's label-free, plus AI on top. So with the data science, we are building in silicon markers, we basically generate the digital twin of the patient, and then we compare this digital twin against other people. So basically we compare people. And so far, we have digitized about 240,000 individuals. First, we have digitized healthy individuals because it's important to understand how does healthy look like in women and men, in younger and older people. So we basically have a health baseline from age 18 to 70 plus. We understand how does health look like, and then, of course, we look at many, many diseases. But first, let's watch a movie.
Ali Tinazli 04:26
Life spin only needs 0.5 milliliters of blood and takes only one measurement for all tests. The NMR instrument is the interface between the physical and digital world. The patient's metabolic information is being captured in one single scan with radio frequencies and compared to billions of data points from individual metabolic profiles within moments.
Ali Tinazli 04:59
Sample partitioning for accessing biomarkers and performing health tests is based solely on algorithms. Life spin. Take your health in your hands.
Ali Tinazli 05:14
That's basically how our life spin workflow looks like. We are collecting blood, then we scan these blood samples. One scan takes a couple of minutes. Then the raw data, the spectra, are being uploaded to our cloud. One patient file is 10 megabytes. And then we use basically our algorithms to run again the general whole metabolome. Then along with the patient metadata, we generate the digital twin. Then we use it further for downstream in silico analysis. And then you get different types of outputs in terms of health reports. Now what exactly do we do? So we look at the spectra. These are the raw data, so-called. And then we basically built an AI engine on top of that, which understands the language of spectra with the raw data in the context of health, and then we basically build our own models, and basically refer the different spectra, the different spec information, to the health status of the individual. So, and not only that, we are building our own generative model, but it's not ChatGPT or anything like that. So we built, we built basically our own generative model because ChatGPT understands the language of human beings; our model understands the language of spectra, and this basically, with hundreds of thousands of patient data, is used to fine-tune models to recognize different health conditions. Here, for example, you will see a typical health report from a patient, then you basically see a strong indication for different diseases. Now, diabetes is a nice example because everyone knows about diabetes, and we do a similar disease scan for different types of indications in the areas of oncology and neurology, and then we basically get more further refined information. What we are building is basically software; we are a data company. As we go as an evergreen program, we keep digitizing human profiles. We enter into the commercial phase now in Q4 with hospitals in Asia Pacific. So ultimately, we are aiming to build an online Intelligent Health software which can basically determine all sorts of typical health conditions. And the question now to the audience is, how much do you think it costs to digitize one patient sample to get the information I presented? Please, some numbers.
Ali Tinazli 07:40
Okay, any other guesses? So how much to pay for that? Yeah, 10 to 15, that's about the cost to digitize the human sample. We put cost in terms of software as a service, but we don't want to be too greedy. The objective is to make this really accessible on a population health level so that we can have really affordable information to enable personalized medicine. We have, for example, started a study in Madagascar that we want to show that we can use AI really to lower healthcare costs in terms of personalization because we are addressing topics with high prevalence in the Global South, specific women's health conditions which can be actually taken care of. And in terms of this model, the providers get us the blood or the raw data. We turn this into the digital twin, which results in the precision health information. Then one data set we can use for different types of diseases, like an in-app purchase on your phone. Then we basically partner globally with hospitals that have metabolomics units. So I think we have 10 seconds left. So I'm an HP former Sony HP guy. We have our founders here. I joined as CEO to basically further take the company forward. Jim Woffman, he's on our board; he won a Nobel Prize about 10 years ago. Then we have further people from the industry, and that's pretty much it. So if you want to contact me, here's my WhatsApp. Thank you very much.
