The Melanie Avalon Biohacking Podcast Episode #252 - Nathan Price, PhD and Leroy Hood, MD, PhD

Direct to consumer testing

AI's role in healthcare diagnostics

pheomonhealth.org

Dr. Hood and Dr. Price's current work

TRANSCRIPT

(Note: This is generated by AI with 98% accuracy. However, any errors may cause unintended changes in meaning.)


Melanie Avalon:
Friends, welcome back to the show. I am so incredibly excited about the conversation I am about to have. This is one of those shows that I've just been thinking about constantly ever since I first was exposed to the book and these authors, researchers, amazing, incredible people that I'm going to introduce to you. Backstory on this conversation, as you guys know, I'm really, really interested and obsessed with the science of longevity and the role of healthcare and the medical system and the future of all of that. I've sort of been haunted by it since, well, actually, I was thinking about this. In kindergarten, I asked the school if I could write a book report. There was not a book report required, but I just asked if I could. Then I picked Gregor Mendel to write the book report on. I have no idea why when I was in kindergarten. I've been fascinated with genes and genetics for a very long time. Then I subsequently became haunted by longevity and aging. Really, a lot of that led to the genesis of this show. Recently, or ish, when some publishers or publicists reached out to me about a book called The Age of Scientific Wellness, Why the Future of Medicine is Personalized, Predictive, Data Rich, and In Your Hands, I was an immediate yes. I didn't even know who the authors were yet. Then I found out. A little bit about the book and the authors, the guests who I'm here with here today. This book, it dives deep into the history, the present condition, and the future of healthcare, where all of that is going. I mentioned genetics, the role of genetics, and not just the genome, but the proteome and everything beyond that. Specific issues like Alzheimer's and cancer and the role of AI and biohacking, although I don't know if they actually call it biohacking in the book, but wearables and basically patients taking health into their own hands and having agency. It's just an incredible, fantastic book. I cannot recommend everybody read it. Then the authors behind it, it is by Leroy Hood and Nathan Price. Lee, this man, I'm just so blown away. He developed so much technology, including DNA sequencing technology that literally made possible the Human Genome Project and is just foundational in so much of the work we're doing today with genetics. He's the co -founder of the Institute for Systems Biology. He's won so many prizes, so the Kyoto Prize, the Lasker Award, the Heinz Award, the National Medal of Science. He is in all three national academies of science, medicine, engineering, and science. There's pictures of him with Obama giving him an award. This is crazy. Then we're also here with Nathan Price, his co -author. He is the chief science officer of Thorn Health Tech. He was selected as an emerging leader in health and medicine by the National Academy of Medicine. He's received the Grace A. Goldsmith Award for his work on scientific wellness, and he has co -authored over 200 peer -reviewed scientific publications. What really blows my mind that I just learned is that they actually wrote this book with everything that they're doing. I was like, maybe they got a ghostwriter to do it, but no, they actually wrote this book, and it's absolutely incredible. Cannot recommend it enough. I have so many questions, so Lee and Nathan, thank you so much for being here.

Nathan Price:
It's a pleasure. Thanks, Melanie. Great to be with you.

Melanie Avalon:
To get things started, respectively, each of you, and I know it's a long story for both of you, but can you tell each of you a little bit about your personal story and when you first became interested in longevity? I'm curious about that and then I'm definitely curious how you guys ultimately met to collaborate for this book. you

Leroy Hood:
You know, I first became interested in the whole question of health when I was six years old and my younger brother was born with Down syndrome and this was back in the mid -40s and I remember asking the physician at that time, what caused that? Why was he different? And this person had no clue whatsoever and of course it wasn't until many years later that chromosomes and chromosomal abnormalities became the foundation for beginning to understand Down syndrome. But that catalyzed an interest that percolated I think throughout much of my grade school. I went to high school in a very small school in Montana, 146 people in the school and had three of the best teachers I ever had. And they treated me as an individual rather than as a student and gave me a perspective about how I should think about my career in the future. And in fact, one of them asked me to help him my senior year teach biology to sophomores and I taught six or seven courses out of Scientific American. And the last one was on the structure of DNA and that's what really made me realize biology had enormous fascinating dimension was really going to be an exciting opportunity for the future. He also had gone to Caltech and he persuaded me to go to Caltech and it was a terrific decision because I got a good scientific background although nothing in human biology and that sent me to Hopkins to medical school and they had an accelerated program for unusual students and I wanted to go and learn human biology but to actually do research afterwards. So I went then and got my degree at Caltech in molecular immunology and that really introduced me to the enormous complexity of human biology and started me thinking about just what we're going to have to be able to do to decipher that complexity. And the bottom line is I started in 1970 as an assistant professor at Caltech and said to myself, where do I want to go now that I've grown up? One answer was dealing with human complexity and trying to figure out how to decipher it. And this led to a whole series of paradigm changes that I think really not only gave me insight into human complexity but really framed my view of human health completely. So we, for example, I had a lab set up that was very cross -disciplinary in nature and we actually over the next 20 years or so developed four instruments for sequencing and synthesizing both DNA and RNA and proteins and all of those things. And then two years later we did, a few years later we did a couple of more instruments. So the first paradigm change was really bringing engineering to biology and frankly there were many at Caltech who were very skeptical whether engineering belonged to biology. The second paradigm change was getting invited to the first meeting ever in the Human Genome Project in the spring of 1985 because of the automated DNA sequencer and coming to the conclusion with 11 other scientists that A, it was feasible but B, very, very complex kind of task. And what was interesting is we were split 50 -50, the 12 of us, on whether it was a good or a bad idea. Those objecting were worried about big science, swallowing all the resources for small science, which of course did not happen. And then from there I went to the University of Washington with the help of Bill Gates and set up a department that was cross -disciplinary for eight years and it was enormously successful in inventing key software for the Human Genome Project. And actually two of the faculty members, Rudy Aversol and John Yates, actually pioneered the beginnings of the field of proteomics, which is a very important area. But what happened was it became obvious that I wanted to superimpose on top of this beautiful cross -disciplinary department systems biology and and the bureaucracy at the University of Washington, the big state school, made that virtually impossible. So in 2000, I started the Institute for Systems Biology that pioneered these ideas that the way you look at biology should be global, it should be holistic, it should be dynamical, it should be integrative and hierarchical, and those words all mean important things in terms of being able to take digital information and translate it into the biological networks that underline wellness and disease, and many others have succeeded in doing that. And that was a foundation point that early on I met Nathan Price in this endeavor, and in 2014 we set up the first population studies using genomic analyses and phenomic analyses, and we can talk about just what those are in a few minutes. But we looked initially at 108 people drawing bloods and microbiome analyses and digital health, Fitbit, and so forth, every three months for nine months in 2014, and the effort was such a success that Nathan and I then co -founded a company called Aerofil based on this whole idea of scientific or quantitative wellness and how you could use data from each individual to figure out how to optimize their own health, and every individual was different, and every individual needed to, one needed to consider their own unique data in setting the ideal health trajectory out for them and everything. And over a period of four years with Aerofil, we recruited 5 ,000 patients, and each of them had their genomes done and their longitudinal genomes done, and these data have led to some 25 or so different papers now. And, you know, each of the papers has been fascinating because it gives us a different view of actionable possibilities that either allow you to optimize your wellness and or to avoid disease. And, of course, the final stage has come in what our book talked about is the idea that going from a for -profit company with 5 ,000 individuals, we'd like to see the whole world assume this data -driven approach to wellness. And in 2021, I created a nonprofit called Phenome Health, and its mission was to drive data -driven science to the point where we're now advocating aggressively a second human genome project which over a 10 -year period would analyze a million people with their genomes and genomes. And this, in fact, would give us the data we need to be able to, one, show that those million people really had an enormous gain in quality of health care, and two, demonstrate striking cost savings because we could really begin to avoid very powerfully chronic diseases, and we can talk about that at a later point, and three, it would generate the technology that in time would be able to drive data -driven health into the home, and ultimately think not just about doing the mistakes but spreading it to the world. So the two big ideas that would emerge from that are, one, we really see in a 15 to 20 -year period replacing a health care system that's 98% focused on disease care now, with one that's 98% focused on wellness and prevention, and of course will have the ability to manage disease when it escapes and happens. And of course, the second point is being able to take this data -driven health to the entire world and democratize it, and think about being able to do it in Africa and underdeveloped as well as the developed countries. So that's a long -term solution. It opens up hundreds of questions about, how do we do it? But let's hear from Nathan.

Nathan Price:
Yeah, so I can say a little bit about how I got interested in this space. And you asked also how Lee and I met. One of the things I like to say is that I met Lee in 2003 and he met me in 2005. So I'm nearly 40 years younger than Lee. And so he was a very famous scientist when I was a student. And so I had a chance to interface with him very briefly when I was getting my PhD at the University of California in San Diego and had been pretty fascinated by his life story and science, which, you know, he just took you through a fair bit of that. And so I had done well at UCSD and had a faculty job at the University of Illinois before I graduated, but I ended up deferring it for two years to go work with Lee as a postdoc. This was back in 2005. Was there for a couple of years. Then I went to my faculty job at Illinois and ultimately Lee recruited me back to ISB and we got into a lot of these things together. One of the experiences that was interesting from my standpoint and getting into wellness was I had a good friend when I was at Illinois who was kind of going through a bit of a life crisis and he was questioning his life philosophies and was contemplating maybe dropping out of his PhD. And he was just going through a lot of issues kind of trying to search for what in his life was, you know, just not fulfilling him the way that he used to be. And we had a number of these discussions but over time he had gone to a series of doctors and eventually one did a good panel of blood work on him and found something incredibly simple which was that he was anemic. He just, he didn't have enough iron in his blood. So he starts taking some iron pills. A few days later he comes in and he's right as rain, right? He's excited about the world again. He's moving forward and he goes on to have incredible success at MIT, a hedge fund on Wall Street, leading an AI group at Amazon, et cetera, et cetera. And one of the things that was kind of remarkable to think about is that without the proper, like just tiny tweak of his wellness, of just figuring out what was it that was causing this malaise and this problem from a prevention standpoint, that whole incredible future of success and building the beautiful family that he has and all of those things may not have happened. And it came back to something that was so simple that it was kind of mind blowing. In this case, it was the kind of thing that we would find with some of the conventional tools, but how many more things are there that we don't really understand? So I became very interested along with Lee about how we could, in fact, think about our health trajectory much, much earlier on and not wait around until we get into these late and terrible symptoms. And then the flip side of that is that we're all familiar with the tragedy of early deaths and issues that are hidden from our view. Lee and I had a dear friend pass away this weekend a couple of days ago from cancer, another professor at ISB, a model of health and fitness in every way. Vegan does everything that I'm sure would be applauded by lots of health enthusiasts. And yet you have these silent killers and these aspects that are happening all the time. And so it just became so obvious that the whole structure and framework of how we go about the medical enterprise and how we set up hospital systems and the whole structures focus so much on disease and on care after you get these terrible symptoms is just intervening too late. And so, as Lee already alluded to, so I won't go through that, we just became very convinced that we needed to do something that was different and try to understand health in a deep way, which is why we led to the co -leadership that Lee mentioned around the Pioneer 100 project, the company Aerovale and all those things that I'm sure we'll get into. But that was really my entry point into starting to think about wellness in a deep way a little over 10 years ago.

Melanie Avalon:
First of all, my condolences. I'm really sorry about your friend. It's actually making me think of something really interesting, which is also this week. My aunt passed away from pancreatic cancer, which she talked about in the book. And what's interesting about that dichotomy there is, you know, you're speaking about your friend and, you know, he had this seemingly great lifestyle and was still, and still had this tragedy. And my aunt was the opposite. She was kind of like, I mean, you could just look at her and tell she probably had all the markers of metabolic syndrome and all the things. And so I know this is like a dark question to start with, but it makes me wonder. So even if we can detect these and what we can get into what you talk about in the book and what you do with your work with these transition states and stuff, but even if we can find the beginnings of disease and, you know, intervene, the intervention itself, do you think the answer for that is in pharmaceuticals and the medical system and there's a solution or is it in lifestyle? And if it's lifestyle related, doesn't that ultimately lie on the hands of the patient? Like I'm wondering like actual change here. What's the role of both detecting disease versus the actual solution to that?

Nathan Price:
This is an area that I think is really important for what we call scientific wellness. The type of intervention that you would do in an individual that is not, that doesn't have a terrible disease is obviously very different than what you would do in late -stage disease. Late -stage cancers, we do things like chemotherapy and radiation, right, that are in and of themselves quite harmful, but you do it because you're trying to kill off that, the immediate life -threatening aspect of that tumor, for example. You would never do such an intervention, obviously, in a person that's more or less healthy, but on a trajectory. So that does lead us to more interventions that have to be rooted in things that are very safe. So lifestyle is a part of that, and I think it's an important part. I don't think it ends there, though. I think that we also have to understand how to nudge these kind of networks in ways where we can have an effect. So natural products is an area that I'm very interested in. Obviously, I'm chief science officer at Thorne, so we have a lot of products in that kind of space, as do many others, but I think that supplements, natural products, things of that nature, have a role when you do them early on and you're trying to have an effect on a particular network or a particular process that may be going awry in its earliest stages. I think that's important. Drugs, I think, can play a role if they have enough safety. I mean, we're gonna be learning a lot over the, right now, obviously, GLP ones are hugely in the news. It's been argued they're the biggest trend in the world at the moment, not just health trend. Some of those things that might be helpful for people with obesity, those can be interventions as long as they're safe enough. So I think we can take a breath because it's not probably enough to just focus on just behave better. We all kind of know things that we should do, but I think what we really want to do is to understand these processes in enough depth that we can help people, even if they're going through a lifestyle change, like give them benefits such that they get more bang for their buck, more benefit as you go forward, and I think that can come from lifestyle, from supplements, the natural world, and from pharmaceuticals in certain cases if they pass the right kind of safety standards.

Melanie Avalon:
Okay, gotcha. Yeah, that's a really, really important distinction. And just as far as understanding, so there's a major theme, we've touched on it already in this interview and in the book, this whole theme of looking at the parts versus the whole and, you know, Lee really pioneered this systems approach to health and wellness where you're bringing in this much broader view of everything. Also, side note, Lee, I was, I was really fascinated hearing about your experience in high school. I kind of can see a theme in your life about going beyond just one individual area. You did everything in high school. I wrote it down. You did, you were in the band, you were the co -editor of the yearbook, you acted in place, you were in the debate team and you were the undefeated quarterback. I was like, how is he doing all these things? But it's kind of speaking to this theme of looking beyond any one area. The question I have, because I recently had a guest on the show, Christian Madsbier, his name is Danish, that's why I pause when I say it. But he wrote a book called Look, and it's all about how we see the world. I'm wondering, in order to have a systems view, what is the role of understanding the individual parts first individually? Like, could we have started with the systems approach from the beginning of medicine? Or did we have to come from more of a siloed parts approach? And then when we are collecting all of this data, I mean, there's so much potential data. You talk about all the potential data points and permutations and all the things. So how do you even know what you're looking for if you don't know what you're looking for? It's just overwhelming a little bit.

Leroy Hood:
You know, the simple way to think about systems biology is to say, suppose you had to figure out how a car worked. You could take any individual part of it, say a piece of the carburetor, and study it your entire life and not have any idea about what role it played in the overall thing. So the aspects of what you need in systems biology is first, to have the ability to identify all the components, and then second, to have the ability to understand how the components are interrelated, and then third, to understand as the car moves, how do these components work dynamically with respect to one another, and then finally, you have to take the global holistic, how does all of this come together to give you a functional car? And, you know, it's somewhat of a simplification, but it's very similar for human beings. So in the beginning, we just had to develop the tools to look at individual parts, and part of what I did was just to develop much, much more effective tools, and in some tools it could do the parts analysis very, very rapidly. And so I think the whole idea of a systems approach is to try and understand any part in the context of the operating system, and that's exactly what we're doing with humans. For example, when we looked at, in Arivale, 500 blood analytes, they were interesting because blood is the ideal window for looking at health and disease in humans because it obeys all organs. These organs secrete molecules that are informational, RNA and DNA and proteins and lipids and so forth. And if you can learn how to read those molecules, and we can with the subset of them, then you can make inferences about the health status of your 25 or 30 major organs. So that is a kind of taking a systems diagnostic view of being able to do very broad screens and with even just a component or two in a very complex process, see something is wrong with the liver in the context of these metabolic activities and so forth. And then, of course, once you've localized things, you can look at the biological networks that are a part of that organ in a functional sense, and you can begin to map changes and the dynamics and so forth into those. And that gives you then this integrative and hierarchical understanding of the organ and beginning to understand the role the organ plays in the context of the whole body and everything. So I think this view of systems biology had to wait until we had the tools to begin to look at lots of different elements in the body. But once we had them, then we could use these tools very effectively. And one of the really exciting areas in biology today is understanding the dynamics of what happens in three dimensions and tissues, and spatial biology is what it's called. And again, that gives us a view of physiology and disease that is a whole different opportunity to begin understanding the context in which the disease process is actually operating and so forth. So I think systems biology has been really important. But I will make one point with regard to patient responsibility. In the early 2000s, I first proposed this idea of P4 medicine. It should be P4 healthcare, really, it's come to be. It should be predictive, preventive, personalized, and participatory. And the first three are all science, prediction, prevention, and personalization. And that's exactly what we practiced and was derived from, in a proof of principle way, very nicely, the Irritable population. But his fourth P is really in many ways the most difficult of all the P's because it's a question of how you convince patients, physicians, healthcare leaders, healthcare technology leaders, data scientists, politicians, regulators. How do we convince people that a data -driven health leading to wellness and prevention should actually replace this disease -oriented health, which is how the healthcare system makes its money now, or at least 90's, 70's, 80's. They make money off patients being sick and that certainly doesn't align with patients being well because for wellness and prevention there's very little incentive to really put substantial resources into those things. So I think this fourth P is education, is psychology, is sociology, is economics. How do we deal with all of these different aspects in fundamentally changing the healthcare system from A to B?

Melanie Avalon:
Going back to that testing you were talking about with the 500 blood analytes, I find it really interesting. So like a common blood panel has 14 blood markers, I believe. Were those just, I mean, I don't want to say random, but like how were those decided? Are those telling a common blood panel?

Leroy Hood:
Well, the clinical chemistries in common blood panels have been specially selected to each reveal different things about diabetes, or heart disease, or Alzheimer's, or whatever. So they're particularly informative, the clinical chemistries. And I think in Irvail, we did more than, if I remember correctly, Nathan, more than 100 clinical chemistries. Yeah, we did. They were the basis for many of the actionable possibilities that we could then transmit on to patients that had the deficiencies they marked.

Nathan Price:
and so forth. It's probably worth noting how different our objectives were, I guess, than when you get your common panel from a doctor. Typically those are kept pretty tight because they're looking for, again, a specific disease diagnosis. And obviously you're also trying to keep costs down by doing a small number of molecules. Whereas when we were really going through this, because we had such a desire to try to understand much more broadly what was happening, we used quite a large set of these clinical labs that are well understood so that we could tie them back to something that we understand that ties into certain processes, and then use that as like a Rosetta Stone to try to understand what was going on in these much, much larger data sets that you can generate with what's called omics data, all the metabolites or all the proteins or all the genes or whatever. And so that ended up being, and those can actually be run fairly cost -effectively. So there's a pretty strong argument that as we go forward, we'll be able to transform a lot of these assays into much more data -rich kinds of measurements and be able to do that cost -effectively. But the medical community is not quite there yet. There's a lot of validation still to go.

Melanie Avalon:
So when you are trying to determine which markers to look at and making sense of all this data, do you find that the more you find, the more you need to find? And what I mean by that is you reference this even larger data set that you can analyze via finding a few less markers that you actually test in the person's blood, if I'm understanding it correctly. My question is, the more we learn, do we find that we can test things that will indicate more information because the program or the data indicates that the data goes together? It might show up in a more singular marker compared to all the data that would otherwise be needed. Or do you find that the more you research, the more you need to know and to learn? And the example I'm thinking of in my head is I think I currently have partnered or worked with three different biological age companies and they all have completely different formulas for how they determine biological age. And that's a little bit confusing to me. So yeah, how do you make sense or decide what to test?

Nathan Price:
So one, when you, when you analyze the data, and I'll come back to biological age in a second, but when you, when you analyze the data, yes, you're able to find out new things off of the larger data sets that you measured that are relevant to health. So for example, one of the studies that we did looked at a common drug, right, that people take statins, right? So millions of people are on statins to lower LDL cholesterol, for example. So what the data showed was that, one, the efficacy of the statin and how much lowering of LDL cholesterol you would expect was very strongly modified by the microbiome. So we were able to discover a pattern in the microbiome that would, that was, you know, strongly associated with whether or not, you know, you would have like a lot of lowering or not. A big side effect of statins is that you can get an increase in the rate of diabetes, there's a 9% increase in diabetes. But it turned out that only people that had a certain type of microbiome would actually showed evidence that the markers for diabetes would jump in them. Other people with different types of microbiomes would not. We could show that, that the lowering of LDL cholesterol by lifestyle intervention was predictable from your genome. Turns out there's another pattern in the genome that predicts whether or not when you, if you're trying to lower LDL by just doing lifestyle and not getting on drugs, whether that's likely to work. You can predict that. So there's many, many things that come out of these kinds of studies. And that's just the tip of the iceberg that will be predictive. The other part of your question around biological age, that is exactly right that there are many different ways that you can measure biological age. You can do it from epigenetics, you can do it from metabolomics, you can do it from clinical labs. And so those are all telling quite different types of information. So I do think it's a real question of whether or not, there's probably not like a universal biological age. There's a biological age is associated with different organ systems. And so there's kind of a multiplicity of ways to think about that. Lee, you probably have more to say on that as well.

Leroy Hood:
Well, yeah, I think the really important thing about biological age, that is, the age your body sets you are as opposed to what your birthday sets you are, is, of course, the lower the biological age is with reference to your chronologic age, the better your aging. So for me, one of the really important criteria for these assays is can they make recommendations about how to optimize that biological age. And from the Erville population, we were able to show that you could actually calculate biological age using all of the different classes of blood analytes, clinical chemistries and proteins and metabolites and so forth. So in time, we actually licensed this ability to use these analytes to do biological age to thorn. And one thing thorn found attractive about it is it did have intrinsically within it, inferences about how you for the individual could actually optimize your own age. And it actually allowed you to beautifully determine the biological age of major organs like the kidney and liver and immune system and metabolic system and things like that. So I think these things measure slightly different things but the second aspect is do they give you information on where to go next in optimizing your aging? And I think that's a very important part of the issue.

Melanie Avalon:
The information that showed a person's ability to lower their LDL with lifestyle, was that LP little a or was it something different?

Nathan Price:
It's something different. So what we looked at there was a what's called a polygenic risk score. So everyone has these little chain differences in their genome, right at certain locations or base pairs. And if you sum up a whole bunch of those, you can get you can make a prediction just from a person's genome of what the likely level of LDL cholesterol is in their blood. That's not knowing anything about their lifestyle, their diet, how much they exercise, nothing. There's there's a genetic component. And so what we found was that if you looked at a person's actual blood measure for say LDL cholesterol, and then the genetic prediction for their LDL cholesterol, the bigger the gap. So if your genome predicted you were low and you were high, those people were very successful at being able to lower their LDL cholesterol by by lifestyle. If their genome predicted high and they were high, we saw no statistically significant ability to lower LDL cholesterol. This is tested across about 3500 people. And so, so that so, and it wasn't just true for LDL cholesterol. This was true for raising HDL cholesterol. It looks like it's true for things like improving your hemoglobin A1c, the biggest, you know, one of the most used markers for diabetes, or pre diabetes. And so this is if you think about it, an entirely new category of variable that we should be using across all of medicine, which is the difference between your genetic prediction and the actual value. And it gives you a map to all the things that are likely to be changeable in you for the, you know, for the least amount of effort. So as we start talking about, you know, like return on health investment, or we're trying to do interventions for people early, one of the things to do is actually this map of your genetics, which will, which will tell you what you're most at risk for. And that's that everyone knows. But the new thing that we found was that it also tells you the things that are most likely to be changeable for you by your lifestyle. And that that that we think is a really big deal for precision approaches to prevention.

Melanie Avalon:
That is so cool. And now I'm just trying to like quickly do math in my head. So in that model, would it be the same amount of probabilities for, so a person's baseline state. So say you have two different types of types of people. One person who's following like a super healthy lifestyle and another person who's just doing presumably everything that you could be doing wrong. Would both of those cases, you get the same information as to their disease risk for all the different things or would that.

Nathan Price:
change? From the genetic prediction, they'll be the same. So we should separate out like what is your actual risk versus what is your genetic risk. So genetic risk is just calculated from your genome and it's a baseline. Obviously everything that you do in your lifestyle makes a difference. This is why we talk so much about genomics and a word that's not very commonly known but phenomics, which is this idea of all these different kind of measures you can make out of the blood or other tissues. All of that other information is changeable. It reflects what's happening in your lifestyle and your diet and your trajectory and your aging. So it's the genome and the phenome that make a difference.

Leroy Hood:
Let me give you an example, Nathan, that I think was really striking. One of my colleagues at ISB had a terrible family history for cardiovascular disease. And in fact, his father, his grandfather, and his great grandfather had all died of heart attacks or something similar by the time they were 50. So he was a terrible candidate for cardiovascular disease. What he started doing in his mid teens was tracking his LDL cholesterol, which is a proxy for heart disease. And what he showed beautifully is it was utterly baseline level. Up until a point, he was about 35. And then all of a sudden it shot up dramatically. And he was athletic, he was lean, he was on a good diet. None of those things had changed. It was something in his body that triggered that process. And the only thing he could use to bring it down was a set. None of the lifestyle approaches even touched it. So in a sense, there are two dimensions to using these polygenic markers and assessing genetic risk. One is whether you have the risk. And the second is do you have the analytes that are characteristic actually of people who have the disease, the highest risk people seem to have the same analytes as those that have already transitioned to disease. So one of the cool things is we can watch these people and the analytes that would indicate they're approaching a transition point to that disease and we can treat them preventively ahead of time. And there are now hundreds of polygenic scores and there are probably 10 or 12 that are really very good for many of the most common kinds of diseases. And this is a way that we may really be able to effectively prevent transition from very high risk into actual bad but normal into actual disease.

Melanie Avalon:
And actually, so zooming in onto that transition concept to disease, I guess a really important question is, how do we even define what disease is and when a transition occurs? And are the current markers for diseases today, are they a little, are they arbitrary? I just don't know how you actually define when disease starts.

Leroy Hood:
Well, that's really a complicated question because there are some diseases that are caused by one gene, you know, sickle cell anemia is an example and there are a whole series of other ones and clearly the disease emanates from that single gene and so the whole focus is on how can we correct the deficiencies of the gene or more recently, how can we really design the gene so it's normal by genome engineering and so forth and you've heard a lot about that for single gene defects and so forth and I think the prediction is and then next 10 or 15 years we'll be able to correct pretty effectively a lot of single gene defects once we work out the details. But the second category of disease is the one that Nathan indicated. That is, it's a whole series of genes that each contribute a little bit to the disease and so forth and there the most effective ways to begin to understand it is to look in terms of disease -perturbed biological networks and the organs in which they reside and the recommendations and I think one possibility in this regard that's very interesting is Alzheimer's disease and Nathan may talk later about a digital twin for brain health that has had remarkable predictive powers for Alzheimer's but so multigenic is another possibility but then there are other diseases that come about as a result of your behavior, you know, if you only eat McDonald's, hamburgers, you almost certainly are going to end up both obese and diabetic and environmental triggers, toxins, poisons, black mold is can for some people be an enormous inductive agent for Alzheimer's disease and that's really been demonstrated beautifully. So disease, you have to define where it's coming from and then you can begin to understand it in the context of gene deficiencies or behavioral deficiencies or environmental deficiencies.

Melanie Avalon:
Is Alzheimer's possibly, I mean, is it one of the biggest examples we have of, I guess, failures in the clinical trials history of pharmaceuticals and all the things?

Leroy Hood:
It's the worst example by far. For 15 years, people of pharma has plugged away at amyloid and tau and things like that, which almost certainly are consequences of rather than causes of. But Nathan, do you want to talk about what you've done with the digital print models?

Nathan Price:
Yeah, I think, yeah, Melanie, I think it's a really good example for a number of reasons. So one is that it's a very strong example of why prevention is better than late stage treatment. Once your brain and you've lost a lot of your neurons have died and you've lost a lot of your synapses, the notion that you could take like a small molecule pill and like recreate those connections in your brain, it's totally fanciful, right? There's no chance of being able to do that in some significant way. But if you're thinking about it from the standpoint of prevention, not letting those neurons die in the first place, that's a much, much easier problem, right? And we wrote a piece on this for the LA Times last year, if people want to check that out, around the need to invest more on the prevention side. Of course, that's not a new topic or not really a new idea. It's something we've all known since we were three years old. It's basically the story of Humpty Dumpty, right? It's easier to keep it from breaking than it is to try to put it back together once it's, you know, once it's scattered, which is what happens to the people's brains when they're late stage Alzheimer's disease. So one of the things that we've been very, well, I'm going to say one more thing. So I got a chance to give a talk on the digital twin models as one of three keynotes at the NIH meeting on combinatorial therapies for Alzheimer's disease a few months ago. And it was really striking. So the first talk was a pharma talk, very expertly done, really good, great scientist, but it was hard not to be struck by how many, you know, like the whole pipeline, but how much money has gone into this and how small the effect sizes are of the small number of drugs that even show any kind of effect. And we failed about 450 clinical trials so far, mostly around amyloid. Then the second one was on lifestyle intervention for prevention. And for lifestyle intervention on prevention there, there are massive effect sizes, big things you can see over the course of years by just changing lifestyle. And it's different for different populations. The number that really stood out to me from that one was that the Hispanic population in particular was estimated like 57% of Alzheimer's could be avoided by lifestyle interventions aimed at prevention. So it was so stark to go into that and see when you're talking about prevention, you're talking about years of healthy brain life. And when you're talking about treatment, you're talking about slowing the negative, you know, the decline of the progressive decline by a little bit for a few months. And you're also talking about a cost on the ladder that is tens of thousands of dollars and the lifestyle is free to very cheap. Now, the digital twins that we've alluded to where this comes in, and this was work that we did with Tom Patterson and a company called Embody Bio. And we worked together with them closely from Thorn for the last three years on building out these digital twin models. But basically what these do is it lets you map a trajectory. So given a set of blood measures and some genetics and some information about lifestyle, you can build a model that will simulate how the brain maintains health, uses data from about 1200 different papers. It's a complex physiologic model. So we're simulating how the brain stays alive, basically. But what that lets you do then is take those information, you can do a forecast of the most likely age of onset for Alzheimer's for somebody. There's a probability distribution associated with that, you know, so there's some uncertainty. But then you can prescribe personalized interventions and then calculate the degree to which it shifts that whole probability of when you might get Alzheimer's into the future. And you can see the expected benefit that would be predicted from any different kind of intervention strategy. And so this is a really interesting capability. And it leads us to a whole series of opportunities around doing clinical trials in new ways, which we could talk about, providing evidence for, you know, these kind of multimodal combination therapeutic approaches. But it gives people an ability to navigate and have some sense for not only where they're at, but what they can do to try to preserve brain health for as long as possible. And so those are really incredible capabilities, I think, that are now online.

Melanie Avalon:
Okay, I have a question about this, and I was wondering this all throughout the book with these digital twins. So is the way the model works, is it creating a model that is mechanistically functioning a certain way, and then you're putting in these different factors and seeing how it affects the timeline? Or are you putting in the factors and it's looking at data? It's like looking at patterns of other people to see how the factors would affect the timeline. And the reason I'm wondering that is because just from my naive perspective, so many people talk about how we don't know what causes Alzheimer's and we don't know what's happening. So how can we have a model that can be functioning with a mechanism if we don't know the mechanism?

Nathan Price:
Yeah, so that's a great question. So the answer is it is the former. So we are making a calculation. So what the model is asserting is that we believe that more or less that this mechanism that we put into the model, what we show is that when we, and I'll talk about what that is in just a second, when we look at this approach, we're able to match a huge amount of the data that's out there across 30 different human clinical trials and research studies. And so we are positing. So what we did over the last three years was basically to build a detailed mechanistic model of what we think is going on in Alzheimer's. And what we find is that when we simulate with that model, and we compare it to data in the real world, that it matches really well. So that's not a proof, right? So I wouldn't come out and say like, oh, we've proven that. But it has been a remarkable experience for me, I have to say, to where I feel like there is a reasonable structure that I think is at least getting, you know, a lot closer to what's going on. So, so the way that we're thinking about this, so one is, we really view, and one of the big triggers for dimensions is basically centered around energetics in the brain, keeping enough energy to keep your neurons alive. So one of the things that happens is, well, your brain is a huge energy hog, right? It consumes 20% of your body's energy, and it's 2% of the biomass. So it's 10 times more metabolically active than an average tissue. So you've got to feed energy up to this thing all the time in order to keep it alive. Now, one of the things that happens as you get older is that your ability to perfuse oxygen into your brain goes down. It's helped by exercise to an extent, but it goes down. And as that goes down, it becomes harder and harder to keep neurons alive. And that's not, and that's spatially, and it depends on where you're at in the brain, okay? So it's not uniform, right? It's what we call heterogeneous, but it's a pattern. So as you start having certain neurons that are getting into that lower oxygen condition, then some other genes start to really matter. So there's a variant of APOE called APOE4, which gives higher risk for Alzheimer's. And there's a variant, APOE2, that gives lower risk, and APOE3 is kind of in the middle. So it turns out that when you need to keep energy generation efficient under low oxygen conditions, you really want to keep cholesterol levels and astrocytes low. Now, APOE4, APOE has a role in the transport of cholesterol out of astrocytes. APOE4 does it very slowly, so that concentration stays high. APOE2 does it really fast, so the concentration stays low. And what that means is that people with APOE2 have much more energy efficient production in their neurons under low oxygen conditions. When we simulate those two facts against a backdrop of, you know, we do 10 million digital twins, and we compare that against the population of the United States, it predicts very closely the age of onset for Alzheimer's for all the different genotypes, whether someone is APOE33 or 34 or 24 or 23 or whatever. And we find that that really closely matches. We then build out from that kind of central piece, and we just build out a model that tracks the effects energetically of all kinds of different aspects of Alzheimer's, which is like microglia that are clearing out debris from your brain effects. I'll come to amyloid in a second, but you basically build this whole thing up. And then when you look at that related to all these other risk factors that are known in the literature, we find that we're able to predict those risk ratios quite well also. And there's a figure, you know, I think we had an early version of this in the book itself that shows kind of how that goes. So, you know, so still work to be done. But what the digital twin model is positing is it's saying if we build this mechanistic model, and we say that we think this is the real driver for Alzheimer's, what we find is that it explains a lot of the data that's out there. And so this is and none of the pieces are new, all the pieces are known. But we just took all those pieces, again, working with Tom Patterson and embody, but basically put those pieces together. And, you know, and then simulate that as a whole. Now, another thing that emerges from that is a view around amyloid. So amyloid has been the target for most of these clinical trials that have failed. And one of the things that comes out is that Once your neurons start dying, you start losing synapses. You need a certain amount of background synapse firing to do what's called heavy and learning, which is what fires together wires together. And when you start losing those synapses, you can't do that anymore. So your brain has to secrete a molecule in order to recruit additional synapses so that it can continue doing cognition. And what is that molecule? Amyloid beta, amyloid. So when you give a drug and you stop the production, and you clear all the amyloid out of a brain, what does it immediately start doing? It immediately starts secreting amyloid again, if it still has these underlying metabolic deficits. And so if that's true, right, and there's good evidence for this, we didn't even discover that, right? We just pulled it out of papers and put it into the model and then simulated what the implications were. But when we do that, what we find is that, what it means then, is that we have wasted somewhere between, there's different estimates that you can find, between 300 billion to a trillion dollars over the last few decades, chasing what is not a cause of Alzheimer's, but is actually part of a compensatory mechanism. And so, and I've heard some people argue that we should do these clinical trials where we just need to do anti -amyloid a lot earlier in people. Well, if our approach is right, that would be very harmful to patients and it would cause more rapid cognitive decline, right? So there are ways to test this. That would be a very bad, very bad outcome. And so it does get a little complicated and there is some slight benefits shown to some recent anti -amyloid therapies, but we think our models actually predict this as well, because one of the things that the amyloids can do, and there's different ways, so our point of view is that the amyloid plaques are not drivers, but amyloid itself can embed in the membranes of blood vessels, which will constrict them. And when you constrict them, you lower that oxygen, you make the energy generation process harder. And when we simulate that quantitatively, we think that's the reason why you get a slight benefit from some of the current anti -amyloid therapies, but of course it comes at great risk because you also have this, a lot of terrible side effects that are caused by them as well. So anyway, I don't wanna get into all the nitty gritty here, but it is a different paradigm from saying, oh, it's these aggregates of proteins that are driving Alzheimer's disease to instead look at it as much more of a metabolic disease. And in fact, a lot of neurodegenerative diseases are being looked at this way. Chris Palmer at Harvard, I had a really lovely conversation with him a few months ago. He's written this other book called Brain Energy. We got into it and into a really interesting discussion and he's coming to a very similar conclusions that about how important these metabolic energetics are behind maybe not even just Alzheimer's, but a huge swath of brain diseases. And of course it's not hard to imagine that's true for the very basic fact I gave at the beginning. The brain's a huge energy hog, problems with that delivery are gonna be endemic across all kinds of different diseases. So that's really how we're coming in on that. But so we are, so it is, the model is basically a complex hypothesis about how does the whole disease work. And that's the structure that I think should be tested rather than these like single compounds. We really wanna be testing kind of these holistic views of things like the digital twin models that are representative of an entire complex hypothesis about how a disease works mechanistically.

Melanie Avalon:
This is all, no pun intended, beyond mind blowing. When you have those digital twins and well, so has the data from them or the results pretty much lined up and if it ever doesn't line up, do you just adjust the model or do you have to start over from scratch?

Nathan Price:
Yeah, you don't have to start over from scratch. So models can be, the nice thing about models is that you're building a complex hypothesis. You disprove them in parts. It's not like the whole thing is wrong. I mean, it's based on a thousand papers. They're not all wrong. So you do have to adjust. Now that said, you always have to be very careful about overfitting. Because obviously if you're just taking all the data and you're just fitting a model to it, that would be pretty meaningless. So typically what we do is that we do take part of the data and then we will calibrate the strength of an effect or something like this in a certain scenario situation. And then you apply it in different situations or different genes or across a wide swath of the population or into new studies. And then you're looking at how well that predictive ability holds up. And then one of the things that I was really working hard towards was how we could get this into clinical trials so we can really test this out. But I think that's gonna move forward and we're figuring out exactly which group to go with, but that will move forward in some form or another.

Leroy Hood:
I will say there's one more twist we should put into the Alzheimer's. The idea that there are different subtypes of Alzheimer's I think is going to be very important because the individuals with the APOE4 -based Alzheimer's are independent from the polygenic markers that have been generated for Alzheimer's. And so they're looking at two different forms and types and we demonstrated that beautifully by taking a three -generation family with 43 different individuals and doing their complete genome analysis so we could get the SNPs and we could get the APOE2, 3, and 4 and we were able to show beautifully in those three generations that the polygenic Alzheimer's was different from the APOE4 Alzheimer's and in fact they segregated independently in the family. So some people had a double whammy of having both genetic factors and others had one or the other or none. So Alzheimer's is still going to be quite a complicated and heterogeneous disease.

Melanie Avalon:
I mean, just that turn with the amyloid is just crazy if it would be all of this work and money and pharmaceuticals and thoughts and theories, if it's all just not accurate. Question about, because you were mentioning, Nathan, moving these digital twins potentially into clinical trials. So a question or a concern I have, maybe not so much with the digital twins, although it might relate, you talk in the book about high risk and low risk individuals for certain diseases and how that can affect who you actually test and use in trials. Is there the potential of pharmaceutical companies using that and basically setting up trials to favor the outcomes they want because they're using certain high risk or low risk individuals?

Nathan Price:
That's a really good point. The answer is definitely yes. Now it's not necessarily a bad thing depending on how it's set up. What I think you're getting at, which would be a bad thing, is if you kind of blind it, you bias the clinical trial to this very narrow population where it works and then you apply it generally across the whole population as if it was going to have the same efficacy. That would be obviously bad. Where I think this could be useful though, and this is where these trials being regulated by the FDA and so forth I think is really important that there's good oversight because now if your use case in the actual world is coming with a biomarker or some sort of measure or a digital twin model or to go to the next level, but with something that will figure out whether or not the person is likely to benefit from the treatment, that would be a big step up. A pharma company, there might be a drug that would otherwise fail in the general population, but if you understand where it's really applied and you understand which part of the population it will really work in, as long as you're applying that in the population, then it's a good thing, but it is a nuanced area for sure.

Leroy Hood:
You know it's really an important question and I can give you an example that a nature paper of six or seven years ago illustrated and that is they looked at drugs the 10 most popular selling drugs in the U .S. today and for each of the drugs they estimated what fraction of the individuals they treated actually had a favorable response and the range of responses generally varied from 1 in 4 to 1 in 25 and on average for the whole thing about 10% of the people responded to the drugs. Now with a phenome approach to population studies we can very quickly get biomarkers that almost certainly will identify the responders for each of those drugs and today we spend way above 600 billion dollars a year on drugs and suppose you could do away with 90% of those drugs merely by having the biomarkers that Nathan talked about tell you for a statin or for whatever this is a subset that's going to work and I think that's going to be a very important part of the data driven cost containment that will come in addition to avoiding transitions to very expensive chronic diseases.

Melanie Avalon:
That chart you have in the book about the 10, I think, yeah, like you said, the 10 common pharmaceuticals and their efficacy is just really fascinating. We can put it in the show notes for listeners. And I kind of just went on a whole like field trip in my head because I was thinking, well, if pharma companies did do this where they used biased participants to get the outcome they wanted, I was like, well, then the drug wouldn't necessarily work on the general population anyways. So it would be a moot point. But no, because it's the point you just made. You know, we have all these pharmaceuticals.

Leroy Hood:
work in a fraction of patients, yeah.

Melanie Avalon:
And they're still the most common pharmaceuticals on the market. So clearly, that's not a safety net for that issue, or at least not right now.

Nathan Price:
Well, and this is why we want to move past, you know, this whole notion of, because right now, right? You do the blockbuster drug, you get it approved because there's some significant effect. You do these massive trials so that you can see that effect and then you apply it to everybody, right? And only which a minority will benefit from. That's the whole notion behind like personalized medicine or precision medicine, where you want to get to where you can get a treatment that's right for that patient. And this is where going forward, I think the digital twin models are going to be so important because you can simulate treatments in advance, right? When you get these things centered and you can ask a question about, you know, if I intervene in this way, in this particular patient, given their particular biology, right? What's the expected outcome? And so, and you can think more about combinations and sets of things that you might do that gets harder, you know, depending on, you know, the more dangerous that the drug, you know, the drug or drugs in question, but especially if you're intervening earlier with safer interventions, you can get really deep into those kinds of questions. And so there is just a huge need for understanding, you know, what's going on in a particular person's body and just a much, much more fine -grained, much more intricate personalized way than we do today.

Melanie Avalon:
and actually still within this world of concerns and applying this data and learning things. So my sister -in -law, she's actually at Northwestern right now, she's in their program for genetic counseling. So this is all, she's just so fascinated with all of this. I was asking her like what concerns or what has she seen working with people and relaying this information about genetic risk to patients and such. And she said one of the concerns she has is that people come in having done like 23andMe or something and it'll only test for potentially like one variant or one version of a risk gene, like BRCA for example, when really there might be other versions. And so she said she's worried about direct -to -consumer tests for people because it might not give the full picture. So what are your thoughts on that? What are your thoughts on people doing direct -to -consumer or getting data like this and then potentially learning their genetic risk but it might not be the full picture. So it could be a false positive or a false negative.

Leroy Hood:
I think the important point to make is genetic risk doesn't at all mean you're going to get the disease. And the general population doesn't understand statistics very well at all. And I'd be worried about giving out risk things without counseling that informs you really what the risk means. And a point we made earlier is these polygenic scores and people with high risk, unless they have the analyte changes that go with the disease transition, they may not be in a position where you have to do anything at all. So I've argued increasingly that with the genetic risk scores, what we should do is give them to the physicians to track the patients and where it becomes appropriate in the blood analyte testing, then to inform them and think about what one might do. But I think just giving these risks out to patients that don't go through any kind of educational process, I think is extremely dangerous.

Nathan Price:
Yeah, I'm gonna weigh in on that too. I'm gonna be a little bit different than Lee on this topic, because I think I'm more liberal on this. It depends exactly, you know, obviously the nature of what you're giving out. And there is, you know, there are definitely things that you can get to in more detail. That said, I think that the people that have tried to argue that there was, you know, actual harm to patients from say, like a 23andMe test or an Ancestry. you know, test is incredibly limited. I've never seen any. I think I think there's also, in my mind, a pretty fundamental right to the access to information if you want it.

Leroy Hood:
I agree with that completely, Nathan. If somebody wants it, absolutely we should get it to them.

Nathan Price:
And so I think that it's important to be educated. I think it's good to say those kind of things. I think that often, though, people can be way too overly paternalistic about the kinds of things. Well, I'll just say in my own state, right, I'm not allowed to order my own test, right, I can't measure my own microbiome, because the state of New York thinks it's I'm incapable of understanding it. So, you know, which really Wow, yeah, which kind of pisses me off. But it's like, you know, so there there are these kind of things that you can that you could we were on with the with the mayor's office a little while ago, they were very supportive, but we'll see if anything happens. So there is a lot of that. And the flip side is you can buy a 23 and me test for 50 bucks. A hospital system will charge you $3 ,000 per gene for many of these things. So yeah, if you're talking about if you're going to make a serious medical decision, yes, work with the hospital system. But if you want to have access to knowing a little something about, you know, what's in your genome, or what, you know, what you might, you know, and what have risk for, or what your ancestry is, or what this or what then you understand kind of what what that is, man, I wouldn't want to live in a in a in a society where we said, Oh, we're going to block people from having, you know, cheap and, you know, cost effective access to the kind of information that's coming out of science if they want to, I mean, it would end all biohacking. So, you know, like, for in a larger, large setting. So, so anyway, so I am I am, there are nuances and balance, like in everything, but I would I would be more on the side of I feel people have a have a right to get information if they want to.

Melanie Avalon:
Like you just said, I mean, this is sort of literally the foundation of the biohacking world. It's people who really wanted to have access to and get this information and, you know, learn and not be completely behind the gates of the conventional medical system. And maybe while we're talking about the concerns, one last concern I have about everything. And so I understand that, I understand chat GPT is not representative of, well, I'm guessing it's not representative of the AI that is happening with all of the analyzing of data and the medical system. But I use chat GPT a lot almost every time I end up getting in an argument with it because I find that it'll like hallucinate. So it'll just make up stuff. And I learned recently why that is that that current model or program can't not know the answer. So if it doesn't know the answer, it just makes up an answer, but it doesn't know what's making up the answer, which is just like highly concerning to me. But I've heard in the medical version of this that there's like black box answers where it'll just spit out answers and we don't know how it got to that answer. So what is the role of AI and the truth that it actually knows? I loved the part of the book where you talked about the role of humans and how one of the benefits of humans, especially like humans coupled with AIs, is humans have a broader perspective where they actually know when you need to look beyond the rules or break the rules. And that risk aspect can actually bring more truth or answers when coupled with AI. But I'm just very concerned based on the arguments that I get in with chat GPT, you should see our transcripts. It's not good. Do you have concerns here though for the medical system side of things?

Nathan Price:
Yeah, I have a lot of thoughts about that. I'll preface it by saying that I find these technologies amazing. I love playing with them. You know, for all their awards, I think they are an incredible beacon for us to get access to, to make information like super accessible. So I'll start with that. There are issues, though, obviously, like the one you brought up is a big one, right, which will hallucinate an answer, right? It has to come up with an answer. That I think is going to get quite a bit better. We've actually been building some things like this, where you can vectorize like responses against curated text, for example, that is and you can you can instruct these things to only give responses if you can find it in like a vetted set of information. So you can really reduce hallucinations quite a lot. The other big element are the agent based AI is now right where you you have, for example, that first AI, like if you're just using chat GPT, obviously, it's solving the problem of it's always just predicting the next word. So it has to write an essay basically starting from a and just going to the end with no stop. And it just plows ahead right. And if it makes a mistake, it just makes a mistake. But you can have agents so that you now have a second AI that comes in and reads the text that came out for the first one, and acts as a critic of it from a scientific standpoint. And you can build agents like that. So, so, you know, without getting into all the details, I think there's a lot of things that can be done to mitigate these kind of issues as as you go forward, such that you can get, you know, better information off of these things. And I do think that's going to be really important. Because at their best, one of the things that I'm excited about is that they give people an ability to get information explained to them at a level and background that's appropriate to them. And so I think that we're going to work through a lot of those kinks. And that'll be good. Now, the one other point I want to make, and then I'm sure Lee wants to come in here too, is that the one big limitation they have, of course, is they have no model of the physical world, they only know text, right? And text is a very small part, right? As humans, we interact with the outside world, you know, we have an understanding of, you know, of the physics and motion and and all of these pieces that essentially don't exist for the LLMs. So there is going to be a large language models of people aren't using that. And so that is where I think there's an incredible intersection going forward for the AI space in biology, which is to take these kind of physiologic models like, you know, like we use behind the digital twins, for example, tie that in with these high dimensional measurements that we've been talking about, and then be able to deliver that information in personalized ways via something like an LLM that's rooted in, you know, these kind of approaches to minimize the hallucinations, I think that's going to be incredibly powerful. And I expect those are going to be built out at like, at massive scale over the next few years, because it's just such a clear pressing need.

Leroy Hood:
I would just add, I agree with what Nathan said, I would just add that what we're playing with is the ability to take an individual's data and translate it into a knowledge graph. And of course, the knowledge graph uses relationships in graphical form for the scientific world. And Google, for example, has a knowledge graph that has billions and billions of nodes. And edges and things like this. And when you lay the specifics of digital information into the knowledge graph, you convert it into language. And that's the language that can be with real constraints that are tied to the medical literature. You put it into the large language model. And that, again, I think is going to be very directive in keeping the large language model on track and giving you back relevant data. But the idea that you can interconnect to the whole world of published data, the individual's kind of data, is really an exciting kind of concept. And it assures one that there's going to be a great deal more relevance in the predictions that then come from a large language model. It's been educated with, you know, 20 ,000 different phenomes and genomes of individuals and so forth. And of course, my hope for the future is that the large language model will be extremely powerful in extracting for each individual a prioritized list of actionable possibilities that can improve wellness or avoid disease. And in partner with, say, family practitioners, can make the family practitioner a superdoc, because in time, it'll cover every field of medicine, and it will make the doc as a partner to the large language model a domain expert for all these different fields, and hence able to treat their patients in ways they could never, ever imagine in the past with this extended domain expertise all large language models.

Melanie Avalon:
It's so incredible, and you do paint this picture in the book, but do you anticipate the computer and AI and everything replacing doctors ultimately, or will it be collaborative?

Leroy Hood:
Now we had really a nice example in the book of, you know, the chess world champion that got defeated by a computer who ended up being a partner with a computer and showed that the two of them could beat any other computers in doing chess. So the creativity and originality and out of the box thinking that the human, together with the number crunching of the computer, is, I think, the kind of partnership we want to aim for.

Melanie Avalon:
I don't know if you guys are Star Trek fans, but I love that even in the original series, they still maintain the country doctor vibe, even partnered with the computer. Absolutely. That was my favorite show growing up. I'm obsessed. Well, thank you guys both so much. This has been beyond incredible. Friends, I cannot recommend enough getting The Age of Scientific Wellness now. It is so eye -opening, so mind -blowing. We didn't even remotely touch on barely anything in the book. It's so incredible. So this has been amazing. How can listeners best follow your work? Can they enroll in any studies that you're doing? What's happening right now?

Leroy Hood:
So I can say for PhenomeHealth .org we have a website and it tells you the essence of what it is and we are going to have places where people that would be interested in participating in studies can register and something about the kind of study. So I'd encourage you to look at the website and explore directly with us if you have questions.

Nathan Price:
Yeah. And for me, I would say I'm pretty active on LinkedIn, you know, a post on there, quite a lot of like new ideas or new things that we're doing. So if people want to follow me there, also on Twitter at ISB, Nathan price, and also for some of the testing and stuff that we do, thorn .com. Th O R N E dot com. People are interested in that.

Melanie Avalon:
Well, listeners know, so I actually launched my own supplement line, but it is highly modelled after Thorne. Listeners know, even the look of it. You guys have been my go -to recommended brand for years. And that says a lot because I'm really, really intense with my criteria. So thank you so much for everything you've done there. This is the last question I ask every single guest on this show. It's just because I realize more and more each day how important mindset is. So what is something that you're grateful for?

Leroy Hood:
You know, I have to say, I'm grateful for the fact that I'm 85 now and I feel like I'm 60, and that I have the energy and the passion and the excitement I've had my entire career to catalyze this final paradigm change that will bring a health care that is all about wellness and the prevention of disease.

Nathan Price:
For me, I guess I'm just looking outside my window at the beautiful trees and everything. I'm just kind of grateful for life. Every day of life is such a gift. Maybe it's just because of the – sorry, the passing of my friend, but it's – yeah, every day is a gift.

Melanie Avalon:
Thank you both so much. I am, I'm so grateful for all of the work that you're doing. It is life -changing. I just, I really can't express enough gratitude and I can't wait to see all the future work that you do and thank you for your time and hopefully we can talk again in the future. So thank you both. I'm just so inspired. You're just really inspired me. I can't thank you enough. So thank you.

Nathan Price:
Okay, bye -bye. Thanks, Melanie. Great to talk to you.

Melanie Avalon:
Bye guys, you too, bye.