Gordon: I want to try and understand your sort of day-to-day academic life, although you are a pretty entrepreneurial academic, to be honest.  But just how that influenced those decisions and the kind of companies you’re interested in, and-and finally  settling on this concept of wellness. Before asking about that, did you ever feel, either in your research capacity or just in your way of thinking, that you’re not just at the cutting edge but perhaps a little bit too early?  

Nathan: That’s absolutely always true. I used to joke about that, that we were never any good at anything because the second we were, we would [laughs] always do the next thing!  As soon as we learned something, we’re like, all right, that’s over; now we’re moving to the next thing [laughs]. It is true ’cause well, my Ph.D. work was on building genome-scale computer models for, uh, metabolism, whether that be humans or microbes. then I became really interested in personalized medicine, and systems biology’s kind of always been there, which is just this notion of trying to understand biology based on all these complex interactions and what you can model in a computer and try to make sense of all this really big data that’s being generated. And so scientifically, eventually I really moved to focus on systems biology. And I did a lot of work in cancers. We did work related to, uh, the diabetes, to pregnancy was a big one, Alzheimer’s emerged as a really big area, microbiome, so I was involved in a really diverse set of-of project areas. Uh, when I was at Illinois my first three years, I put in a bunch of grant proposals. And those grants were on honey bee social behavior, cancers. Uh, we did a geology work  with a geologist at Illinois.  And a lot of senior faculty wondered how that could be rational in any sense, and maybe that — well, maybe it wasn’t. But we could do it because the-the core elements of modeling and systems biology were just applicable. So as long as we worked with someone who was a serious domain expert in whatever it was  you could do that. And it was incredibly fascinating from my standpoint. Then I got really serious about, um, collapsing down what we were doing. And over time, the biggest thing that emerged from that, um, was this area that we called scientific wellness — that Lee and I call scientific wellness. And scientific wellness is also broad but it’s just the notion that the way we view healthcare and how we do studies in healthcare, so medical research itself, it’s all really focused towards late-stage disease, right? We do case controls, and if you have an [oral one] and you’re trying to get a signal, you look at late-stage disease ’cause you’re gonna get the biggest  you know, you’re gonna get the best, smallest P values. You’re gonna get the biggest effects. So you do it, and you do it that way over and over. And I became convinced that we could do a lot better than that, or at least we could supplement that a lot, if instead, we started thinking about, how do we extend health span, of course, you’re at the BUCK, so this is not gonna be a surprise to you [laughs], but how do we get at the roots of, how do you actually elongate health  and study these systems.  And when we really started getting into that, we got a lot of pushback. And it was really interesting how many people were kind of vehemently opposed to this research. It was sort of –  surprising to me because they really felt like this was a waste of time. I didn’t understand the vehemence in some cases because, if I was wasting anyone’s life, it was my life. Like okay  – I’m wasting my life by [unintelligible] [laughs]. I should care. So we started studying well people, right, just –  getting a cohort. This is when we-we did this thing called the Pioneer 100 project. And the idea was, could we learn things from these individuals, even though we weren’t selecting for a particular disease? And I had some trepidation about that, right, ’cause I had so many people that doubted this, and it was a little bit different, But if I fast-forward ten years, we’ve learned an incredible amount by studying these individuals. We ended up starting a company called Aragos. We had about 5,000 people over the course of four years that we went through a program like this.

Gordon: Mm-hmm.

Nathan: And we were able to learn a lot about — we can dive into any of these, but how the microbiome is related into aging, how you can predict the diversity of your gut microbiome from, uh, your metabolome and your blood, uh, how genetics predicts the outcome of lifestyle interventions, whether or not you can lower LDL cholesterol by lifestyle. Turns out to be predictable from the genome. And mappings also of the earliest transitions to disease, which you can do by just looking at genetic signatures. And when you look in people that are asymptomatic, what you find are these really sparse signals but that are highly enriched for things that go on to be causal  as opposed to looking late when you have thousands of changes. But it’s a mixture of cause, consequence, compensatory mechanism, and it’s a big mess to sort out. Anyway, all that to say that I just became really enamored of this. And then we’ve taken that approach then to chronic disease, uh, Alzheimer’s being one of the big ones.

Gordon: The critics you had about this approach, was it similar to what you said, they’re not gonna find anything. They know that the P values are gonna be enormous. Was it sort of technical, um, features of the phenotypes themselves that the people thought how on earth are you gonna find anything?
Nathan:  All of us are in a box to one-one form or another, right? Everything we think about is based on certain assumptions. And just the way that science and medicine is typically done is it’s just case control design, right? Well, there’s two boxes. One is the case control box, which I love case control designs, too, you know, they have their place. But that was like one box. And the other box is we study disease.

Gordon: Yeah.

Nathan: So if you don’t have a disease, and you don’t have a case control, what the hell are you doing, right? [Laughs} But that’s where we had to really start looking at, okay, well, what are the kinds of questions you can ask? And that’s where the genome becomes a really big unlocker because you can say, well, we can stratify every person by high to low risk for all the diseases. And we can use that as a metric to come in and then study, all right, well, what’s different between those sets? Okay, well, now we have a question. Now we can ask something. And then you could go into the data and say, well, what-what’s different about everyone that’s on statins versus not —

Gordon: Yeah.

Nathan: — that you could construct experiments in that way. And then you had all the clinical labs. So we had about a hundred to 150 clinical labs. Those are like a Rosetta Stone. So we could say, well, what’s different about people that have high hemoglobin A1C, a marker for diabetes, or that have, um, you know, or have low amino acids in their blood or have — or-or anything. And what became obvious over time was that, as you start stacking lots and lots of different kinds of data together and having it longitudinally, is that you could ask infinite questions. And it was just like being, uh, a kid in a candy store because because before, when you have to do your case control design, you have a question, and then you think how am I gonna do that. And then you map out a trial or a study. You then go ask the NIH for money or whoever.  It’s this long process, and you have to be very careful about what you’re gonna go after ’cause it’s a — it’s a huge effort to get to, can I ask-ask a question. But when you have these dense longitudinal datasets, it’s not the same, right, as maybe setting up the exact experiment you want. But when you have a question,   I just walk to one of the great [computational] people in my group and I say, “I have this question.  are these things related? Does this predict? Does that –? You know, let me know tomorrow”  because you just had all this data. [Laughs] You could just get the answer. And so the speed at which you could work at on these things was incredible. And I just became a total believer that if you really want to understand any of these diseases, you’ve gotta build these really dense, and by dense I mean stacking lots of different types of data on the same people at the same points in time so that you can see the whole movie. And then as you do that, it just opens up so much. And we probably published, you know, 25 or so, you know, high-profile papers, most of them in the nature journals,  and we still are.  And they just keep — they just keep coming because you can just keep asking — you can ask a thousand questions. You can write a thousand papers out of those things.

Gordon: Do you think you’re an impatient person then?
Nathan:Totally, yes [laughs]. Yes, yes. Unless, you know, the Buck is insanely successful, which I certainly hope it is, we only have a finite period of time.  Now  if you’re gonna get me an extra thousand years, maybe I’ll chill out a little bit more, but [laughs] otherwise, yeah, I want to — I want to be done yesterday.

Gordon: Well, let’s talk about 20 years, first of all.  

Nathan: All right, fair enough.

MUSIC BREAK

Gordon: So I want to talk about your, um, uh, amazing working relationship with-with Lee Hood. Uh, my generation, everyone knows Lee’s name, an absolute, uh, incredible scientist, so many discoveries, uh, an entrepreneurial person. Sounds like the perfect collaborator with you. So what were your-your first links in terms of the research that you were doing, and, uh, and how’s that developing now?

Nathan: Yeah. So I first connected with Lee because, as you mentioned, he’s a, you know, a legend in biotechnology. And I had been familiar with his story. And so when it came around time to think about a postdoc, uh, I just had my advisor reach out to him so that we could, uh, make that connection. And so I still remember that entire email he wrote back, uh, which was, uh, “Price looks great.” That was the entirety of it [laughs]! And then we got on the phone and so I went and got a postdoc from him. And we just really hit it off. Um, you know, it took-took a little while ’cause I was pretty much in awe of Lee when I was first starting and the thing I’d say I really learned from Lee is-is being bold and going after big things because Lee is, uh, as you know, uh, not a timid person. Lee likes to, uh, you know, always-always has to have a grand vision that he’s going for. And I learned a lot from him on that front. And then the two words that people use to describe Lee that I would also use is determined optimism. That’s kind of his mantra through life.  And it’s, uh — he’s a fabulous person that way.

Gordon: Around about what time did you start using the word “wellness”? And-and-and, you know, certainly, please mention your book at this point, which is a very important, uh, landmark in-in the way that we think about healthcare. 

Nathan: Yeah. So the book is “The Age of Scientific Wellness,” which Lee and I published from Harvard Press, uh, earlier this year. We really first started talking about scientific wellness in probably 2013, maybe a little bit before that. But 2013 is when we launched the Pioneer 100 project and we were going around to many places at that point in time and talking with people and getting advice and then it became — and then the thing that made it really solid was the Pioneer 100 project where we said, okay, we’re actually gonna get other data, build it, start building a wellness program. 

Gordon: Was there a feeling that there was a moment in perhaps the Pioneer study where you just thought this is real, this is how we’re gonna make this happen?

Nathan: Yeah, it was — it-it definitely happens, uh, over time. Uh, but there were def — you know, definitely moments in there. Uh, so one, uh, we were de-definitely feeling around and try to figure out what term to use, uh, ’cause wellness was kind of a loaded term.  A lot of people in science didn’t really like it, you know. So we ended up with this — you know, calling it scientific wellness for lack of a bet — we just couldn’t figure out any better term, uh, besides wellness so, uh, to just say that we’re positively focusing on health and not on disease. At the beginning we were really imagining, you know, getting a hundred thousand people and monitoring them over the course of a couple of decades so that we could watch as diseases emerged, so that we could go back and try to figure out like what was happening before then. But one of the eureka moments, at least for me, was when I was sitting around thinking, is there a way we could do that a lot faster. And-and that’s when it-it hit me that we didn’t have to wait to watch these things because of genetic risk profiles. Because we have these polygenic risk scores, we’re at the summation of your genetics of what you’re at risk for,  we already know which — not on a per-person basis, but we know that these groups are gonna get Alzheimer’s at much higher rates than the other group, or cancers or whatever it is because of genetics. And then that was just the realization in the early days that, wait, we don’t have to wait for that! We could just stratify everyone based on genetics and then look at the correlations in these dense dynamic data clouds and start to pull out information that would be relevant presumably to the, you know, prodromal state of disease or at least what’s driving that risk. And when we started doing that, that started to give me a sense of, okay, there’s something here that we can really pull out, and we ran the first analyses. And some of the first data was really intriguing. We took coronary artery disease, and we found out that, of all the proteins we were measuring, there was only one that was correlated with that genetic risk in people that had no symptoms. And it as PCSK9. Well, anti-PCSK9 had just come out as the biggest blockbuster drug in that space. Asthma, there was only one, IL-33. Well, four drug companies had that in late-stage clinical trials as a target –. So you just started, and you just started feeling like I’m — it doesn’t feel like I’m starting to cherry-pick these things. It’s you know, these early signs — well, uh, Alzheimer’s, right? You find the differences in cholesterol. Now fast forward this many years, and we’ve done all these deep dives in digital twins and so forth.  And cholesterol trafficking is, we think, an incredibly important part of Alzheimer’s. So I just think that you get this very different view when you look at people that don’t have symptoms. Actually, when you do case control late, we started to realize, you know, it’s a really messy time to try to sort out what’s going on because if you take, you know, cancer versus normal or, you know, anyway, whatever disease, and you do a bunch of gene expression data, which we’ve done a ton of, metabolomics or proteomics or whatever, you’re gonna find hundreds, maybe thousands of differences, right, that are differentially expressed between the disease and the normal. But trying to sort out from that mess what’s causal, what’s the downstream consequences of what’s causal, and what’s compensatory, right? I’ll argue that amyloid, which we’ve taken as the cause of Alzheimer’s, is mostly a compensatory mechanism   for the brain.

Gordon: Mm-hmm. 

Nathan: But we get these mix-ups, whereas if you look early, the signal — you don’t have this massive signal that you’re trying to understand. They’re very sparse. There’s only a few things going on when you look early.

Gordon: Yeah.

Nathan: And so you can start to build this network idea of the first things are very small; then they get a little bit bigger over time. And when you’re studying people as they’re well, with the genome as the rubric to understand, you know, who’s at risk versus not, you start to fill in more of those early stages. And I just think it-it’s incredibly powerful. So a lot of those kind of moments, and they’re just kind of one after the other, every paper, every study where we started to say, oh, we can really learn off this. And then by the time we were in a year or two, it just became obvious we were going to learn a ton from it.

Gordon: It seems like you identify targets and save half a billion to a billion dollars along the way so that’s very exciting. Um —

Nathan: Well, we hundreds of billions maybe, some estimates as high as a trillion dollars cumulatively over-over the years and decades, on anti-amyloid therapies for Alzheimer’s. 

Gordon: Yes. Uh, so you mentioned digital twins. What are digital twins, please?

Nathan: Yeah, so a digital twin is a representation of an individual patient or individual person’s biology so that at least in our hands so you can do two things: one, predict what’s coming, and second, devise intervention strategies so you can change the predicted future. So we built this — uh, and this is really working with, um, uh-uh, Tom Patterson, who we’ve met — uh, we talk about in-in the book.  Uh, this work was — the very beginnings [was] started at, uh, at ISB. Uh, but the bulk of it has been done, uh, since I moved to Onegevity and then Thorne. And in fact, we have a contract, uh, with them. So we’ve-we’ve built this out as an exclusive partnership for brain health. And so the-the digital twin — so-so what you can do with this — and I do find it remarkable; it’s totally changed my view of what’s going on in-in Alzheimer’s disease. So it starts by building a physiologic model.  So we do this by building, uh, ordinary differential equation models. So you’re-you’re simulating dynamics. So for the techy people in the audience [laughs], they know what those are. Uh, there’s a Bayesian network overlay on top of that to deal with uncertainty. And what the Bayesian network, uh, lets you do, uh, just so — for people that aren’t aware of this, is it lets us use as much or as little data as possible. So what I mean by that is, if I know nothing, the model will tell me the trajectory for brain health and-and into dementia for the average human. But as soon as I tell it, uh, “You’re a woman,” it will switch to that. If I say, “You have two copies of APOE4, it’ll switch to that.  Then if I start adding blood measures, or your vitamin D is low, or you’re, um, you’re sleeping well, or you’re sleeping poorly, or your stress level’s high, as measured by whatever biomarker, usually CRP, but you could do others, and so forth, is that every piece of information that you do, that starts to become more and more of a digital twin of you, right? It starts to represent more of your biology. And what you get out of that is — so if I take measurements then, and we get a set of blood measures, genetics- a questionnaire to get aspects of health, it will generate, it will compute a forecast of brain health, a probabilistic map for the future.  There’s a slide I often show. You know, so we take a man who has, you know, a copy of APOE4, we put in all this data about him, and it would predict that he would get Alzheimer’s most likely at — diagnosed at age 61. So this is a person who’s in pretty bad shape. Uh, and it predicts — but there is uncertainty, but it predicts that with with 99 percent probability, that we would expect him to get Alzheimer’s before age 69. And so you get this map. Now you should think about this a little bit like the ghost of Christmas future, right? If these shadows remain unchanged, given, you know, the trajectory on, this is — this was the expectation. But what it will also do is it will predict all different interventions for what you can do in order to push that that probability mass into the future.  And in his case, we-we predicted that he could get up to six years of benefit by a set of interventions. And these were focused on lifestyle and supplements and things like that.  You can also add drugs in there and-and so forth. But the thing that you’re able to do then though is you can build personalized trajectories and an expectation for how much benefit you might get. Now we’ve compared this against lots of data and all the — you know, out — you know, clinical trials that are out there and so forth, uh, looking to do a prospective clinical trial on this, uh, soon. But basically, it’s a capability to really dive in and understand it from a systems point of view. And then just one more thing is that for most people, for most of these predictions, we show that individual interventions, so any one thing usually doesn’t do anything, or it does very little. But combinations that are personalized are predicted to do a lot. Like you can take four things that do nothing individually, and then you pick a combination and all of a sudden, it predicts, you know, seven years of benefit. And the reasons for that are, you know, not hard to understand. You could imagine if you had a clock, right, that was broken or that’s had three different gears, they’re all broken. If you substitute in one good gear, any of them, it won’t work. You have to do them all. The system has to work  right? If you’re optimizing the regulation of a pathway, but the substrate isn’t there, it’s not useful. And there’s all kinds of dependencies like this. And so when you do the digital twins, you can start to get a sense of what are the set of things that are necessary. 

Gordon: I’m a little scared to have my digital twin constructed in case it’s already passed away. Um, but moving on [laughs] from that, uh —

Nathan: No, we-we’ve actually been looking at that a little bit because that did used to happen in the digital twin simulations, right? Sometimes, someone would be there, and they’d be okay. And you’re like, wow, this really thinks [laughs] you’re on the [threshold].  How is this  possible, right? That could –. So one of the things that we’ve-we’ve-we’ve started to add and study in more detail is how much information there is that you’re cognitively normal at the age that you are because that actually adjusts the prediction quite a lot. So you know, it-it starts selecting you out. And then you can see how much — you know, how much further you-you might likely have.

Gordon: So you know, obviously, artificial intelligence has been around since the ’50s and –. But suddenly, it became this everyday work-a-day tool that obviously, I think, you’re very excited about and, uh, you’re applying it. Was this necessary for this field to-to develop?

Nathan: Yeah, I do think so. I actually don’t think we can — aw, I’ll just say it — I don’t think we can solve biology without AI. I think it’s  too complicated without [those kind] of tools. And solving biology’s a little bit, you know, grandiose. But you know, but we would like to be able to have real control, or we’d love it if a tumor is growing, right, we want to be able to stop it. You know, if-if we’re aging, we want to stop that.  there’s  all kinds of things we want. A loved one’s dying — we don’t want them to. You know, so we — we’re — so — and fundamentally, we really are like — are trying to gain, um, control over the most important part of our environment, which is our body.  And so-so there — I do think AI is-is pretty much essential to that effort. That’s why I think this is gonna be the most exciting century in biology and for many reasons. But one of them is this convergence of big data with AI. So  I do believe if — when we’re thinking about things like precision health or precision medicine, the ability to transmit in just plain language your findings, or make — you know, make this accessible to anyone, or make recommendations, being able to encode that much information and deliver it is an incredible capability. The ability to code without actually having to code, to just be able to tell the computer what you want it to do and have it figure out and code it democratizes access to, you know, building these things out. We’re at the — we’re just in the very infancy, but it is with some dangers, I think we all understand that but [laughs], but it is an exci — it is an exciting, uh, exciting technology. 

MUSIC BREAK

Gordon: So I guess here at The Buck, we think of aging and, you know, we can — we can define that in various ways. But aging is perhaps the biggest challenge to, um, good healthcare outcomes. When  you hear the word aging, or you’ve heard a lot of people talking about it, do you see aging as a special challenge for your approach, or does it just fall into line with other outcomes?

Nathan: I see it as a special challenge because I’m very aligned, I think, with, you know, where The Buck would come out on this, which is if you — if you can make a real impact on aging, it’s much bigger than an effect on any particular disease. And it’s not — you know, it’s not particularly close, right, the economic analysis, right? If you eradicated all cancers, it adds two years to average lifespan, uh, I believe is how the numbers come out. But if you could solve aging, like if we all had the healthcare expenses that we have in our 20s and 30s, healthcare would be totally sustainable. In fact, it wouldn’t even be very big. Now that is not in the economic interest [laughs] of the healthcare system. But it’s massively in the interest of human beings, and all of us to have that. So if you could — you know, because you don’t spend that much on healthcare in your 20s and 30s and-and 40s most of the time, right, and 50s a little more.  And we spend a ton at the end, right? Gawande’s book, Being Immortal, goes into this. But you know, we spend a huge fraction of healthcare extending, you know, bad parts of our life. In fact, um, I’m just gonna share this, uh, anecdote that I-I was going to, uh, give a talk at RPI about a month ago, and I was driving up, and I was listening to a book, uh, not a terribly recent book. It was written by Seneca 2,000 years ago. So it’s a pretty old book by the — by the, uh, Stoics. But he got into a section on — just by coincidence — it’s, uh, it’s a book called “Moral Epistles ,” actually a fabulous book. Uh, but he started talking about longevity. And the way he put it, which I loved, I — you know, better than I ever put it, uh, he said the key is that you want to extend your life and not extend your death. And I thought that is  — exactly right. We want a healthcare system that extends our life. It doesn’t extend our death. The second piece that I just have to add is that he started giving some health advice to his, um, to Lucretius is the — uh, the, uh — his, uh, people that he writes these letters to.  But he started to recommend the new fad, which was periodic fasting [laughs].

Gordon: [Laughs]

Nathan: We’ve moved on to intermittent fasting now 2,000 years later [laughs]. I thought that was, uh — that was just hilarious. So anyway, all that to say that I think our human desires of what we want just don’t change. The technologies change. Our capabilities to do things change. But those fundamental like what — you know, what do you want. And aging research really gets at the heart of what we want, which is an extension of our life, not an extension of our death. And so it’s-it’s — to me it’s, um, probably the most important area, honestly.

Gordon: Okay. Okay, 2,000 years on, uh, Nathan, [laughs] I guess — and-and-and part of this is described in your book, uh, what-what do — what do we do now with the technologies we have right to — right now today?

Nathan: Yeah, so you’d like a 2,000-year update on [laughs] — beyond what, uh, Seneca talked about, we’ve obviously come a very long way in terms of technologies. Uh, so “The Age of Scientific Wellness,” this book that Lee and I wrote, uh, from Harvard Press that’s out in bookstores everywhere, and it is, uh, written to be accessible to a general audience, uh, it’s gonna take — it will explain, uh, this transition in medicine, what we’ve learned by going through all of these studies of, uh, thousands of people as they go through a scientific wellness program, [from] generating an unprecedented level of deep data on these individuals. And we’ll go through a number of cases, uh, including biological aging, uh, hope, uh, in the, uh, fight on cancers, a new paradigm for Alzheimer’s disease, a vision of the future of medicine and so forth. And so if you really want to get engaged in what you can do now, uh, there’re so many things. Uh, so the testing has gotten so much better, whether it’s genome testing or-or dense blood testing, uh, of various kinds. Uh, there’s a lot of — uh, that you can do in the product space. Uh, I’m a big believer in, uh, a lot of the natural products in particular because we’re talking about things that we want to do from a preventive standpoint. Uh, so you know, most drugs are tailored towards later-stage diseases. Uh, so there’s, uh, many things that you can get into, uh, in that space. And I would just encourage people to, uh, you know, be, uh, be active.  Uh, you can leverage some of these new tools with the, uh, large language models and Internet and so forth where it really lets you, uh, get access to a lot of this information and, uh, plans for-for things that can — that can make a-a difference, whether it’s preventing or pushing off into the future dementias, uh, optimizing cardiovascular health, and so many other things.

Gordon: Thank you, Nathan. Thank you for the [book with Lee]. You-you-you paint a exhilarating, exciting world in-in which, uh, we can do things better and we can reduce, uh, human suffering. Um, and thank you for your advocacy. Uh, you do what you do, but you also take time to-to do things like this and, uh, and to-to-to write the book and to-to get the word out that there really is, uh, serious science behind many of the things you just discussed. And, uh, so-so thank you for all that. And finally, thank you for taking time with us today, Nathan. It’s been a real eye opener. I’ve-I’ve learned a ton.  And looking forward to seeing you again.

Nathan: Likewise. Thank you so much, Gordon. And if people want to find me, you can find me on, uh, LinkedIn. I post fairly frequently there, um, as well as on Twitter @ISB, uh, Nathan Price, and, uh, and also, uh, a lot of what I do, uh, ends up at, uh, Thorne.com. 

Gordon: Thank you.