DeSciNYC

DeSciNYC

← Back to videos

Talk & transcript

Decoding Biomarkers for Longevity

Alexander Lozano, MD, PhD

Transcript

Download text

Automatically generated transcript. May contain errors. Select a timestamp to watch that moment on YouTube.

0:03

[Music] [Music] Woo! Let me see. [Music]

0:46

[Music] You know, versus like it makes you more anxious. [Music] I think we're not here.

1:25

[Music] [Music] That's amazing. I love that.

1:48

They're going to pay you for [Music] [Music] I'll ask you to find [Music]

2:44

[Music] Um, just be careful not to touch this. Okay. Yeah. Okay. [Music]

3:26

[Music] [Music] Test test one and two. Test one two. [Music] [Music]

4:07

We're live. [Music]

4:31

The other question is US.

4:49

[Music]

5:06

[Music]

5:16

Hello.

5:32

[Music]

5:41

[Music] Good night. [Music] [Music]

6:11

[Music]

6:22

[Music]

6:39

[Music]

7:05

[Music]

7:19

Oh, we're going to start.

7:33

Guys, be quiet. We're going to start. Oh my god, it's so exciting. In the back. In the back. Quiet. Um, okay. So, it's the first thing is it's like not a great speaker system. So, uh if uh if people are talking, it it doesn't really work. So, um that's the first thing. And the second thing is this always happens and I guarantee you're not going to be able to hear if you're in the back. So if you come here or here or here. So people people over here um should come up because it gets

8:16

you're not going to be able to hear every single time. So if you're just everyone like scoot up like 10 steps and then you'll be able to hear. It'll be much better like even further further further. Everyone just take 10 steps. If you're If you're there, come here. Even more like up here. Yeah. Yeah. Good job. Even even further. What? Yeah. Yeah. Come over here. Like and and here too. You can squeeze around the back. You'll be able to hear much better. And if you and if you do have the urge to um to need to talk um just like go downstairs and it'll be better

8:57

because there are people that want to try and listen. So uh if if you want to talk that's totally fine but just like go downstairs and then come up uh when you're done. Okay, that's all for that. And then thank you guys all for coming. Um it's great to have such a great crowd and um this is DINYC. Uh we have the pictures by David. Um and he'll post them on the Twitter and on the Instagram in a little bit. So that'll be cool. Um and you'll you can see them there. So you can follow on on Instagram and Twitter. I'll leave it up for like five seconds. I still see a couple of phones.

9:38

Uh so we can see the screen. [Music] Yeah. Do you want to just bring it up? You want to put the speaker here? No, it should go further up. Yeah. Yeah. On the sofa. Just put it here. Put it bring it. Bring it. Bring Put it there on the stove. Just It needs to be up. Maybe. Just leave it. Just leave it. Sorry. Sounds good. Yeah. Okay. So, did everyone get the picture? We wanted to get the picture. Okay. So um so who here knows what decentralized science is? We get this question a lot. Decentralized science. Okay, just a

10:21

few people. So I'm going to talk about what is decentralized science. Oh, first I want to talk about what DINYC is. So this is DIN NYC. Um we've had over 10,000 people come to a meetup in one form or another, which is wonderful. Um, it's most of the people here. Who here is a scientist? I'm kind of curious. Oh, wow. Amazing. Who here is like a tech founder? A founder type or founder or tech? Uh, who here Who here is curious about science? Okay. Which things did I miss?

10:58

Science. What? Hate science. Okay. So, you're here to put science on trial. That's good. Okay, cool. Okay, so I was just curious. Oh, if you're curious, we do have all of the talks if you want to hear it at dai.nyc. And this is also being live streamed out to people online, which is cool. So, hello to the live stream people. Um, we've had a lot of people.

11:21

We had Brian Johnson, we had Martin Skrey, we had research hub people, we had uh Robin Hansen, and um it's a growing list and uh it's just cool. So I want to talk about the things we have coming up next. Um we have one on AI and government um especially in New York. So you can RSVP uh here and it's by this guy named uh Alex Bors and he is an assembly member in New York and he's been introducing some regulation um in New York around uh AI and he's going to talk around talk about that. So that'll be really interesting.

12:02

Did everyone get that? Okay, I'll leave it for two more seconds. Okay, we'll send it out if two. Um, and then we have another one on nutrient transport and cell metabolism. And this is by a postoc um, Timothy Kenny, Dr. Kenny, um, who's at, uh, Rockefeller University and he's doing a posttock there. And, uh, it's going to be really, really interesting. And if you want to RSVP for that one, here's the link to do so. Oh, couple more people with their phones.

12:43

Okay. Okay. So, what is decentralized science? Now, I will answer this question. Um so a lot of people have you have done research in like a university setting and you might have realized that when you do research in a university setting um it's sort of very insular right so that only happens to people like the only things that people talk about are the things that like academ academics care about and so it and it takes a long long time for research to escape that and um it's controlled by a select few and like the funding is only done by people who are in government. So decentralized science tries to say like wouldn't it be interesting if instead of

13:23

um you know science being funded by the government and science being done by just in a select few number of people what if instead like you democratized it and sort of made like a Wikipedia or like a Bitcoin equivalent for a university or a re or a scientific research. So that's what um decentralized science is about. Now that's like the goal and you might have questions like like what are some specific things that make decentralized research different from uh traditional research. Um one is the way science is validated. So uh the typical way that science is done is it goes through something called like peer review. So

14:04

you submit your paper to Nature or some journal and you have three um readers who read over the paper and sort of rate it on a scale of 0 to five of if it should be accepted and then it's like accepted or rejected based on that. Now the problem with that is like the the reader or the reviewer has absolutely no incentive um to do a great job at validating the research because they don't make or lose any credibility or money uh if they're correct or if they're incorrect. So one of the things that happens in decentralized science is that um you have um you have uh you have people um you have people who uh want to uh bet on scientific research. So

14:45

instead of just saying like let's give you an example let's look at how how many people here have used poly market. So poly market is like what people were using in the election right? So instead of going to CNN or Fox News, depending on your political beliefs, um you go to Poly Market. And what made Poly Market really cool is that um like if if the people if you were correct in in val in thinking about who's going to win the election, you actually could earn money.

15:11

And that's like really neat. So we want to use the same ideas in in science. So if if you think a hypothesis is going to validate or not, you can actually earn money or lose money, too. And the goal here is is not to earn or lose money, but to have people be more careful, um, when they're assessing science. So, um, one of the things that we're conducting a study on is sleep because sleep is like one of the, uh, most important things people can do to live longer. Brian Johnson talks about it as it helps repair muscles, it helps support the immune system, you can detoxify the brain, it has cognitive enhancements as well. So if you don't

15:51

get enough sleep, for me I feel clouded and I can't think straight. Not exactly, but it definitely helps to get more sleep. And also if you're trying to live a long life, the sleep helps to like give your body a chance to repair. So um we've been I was looking at some of this research that happened some of these people that have been doing sleep research um using wearables and like Sleep 8 and Whoop. So, if I don't know if you saw this tweet, but during the Open AI drama, um the founder of Sleep 8, which is like a bed that sort of cools itself, tweeted that people are sleeping a lot less during the drama of of of um Open AI when Sam Alman was fired because of because of

16:32

all the the tensions. And he noticed this uh a a downgrade in people's sleep during this. And I thought this was really interesting because you know during this is like a distributed trial that he's running and he's able to just look at the data and figure out these sort of interesting characteristics. And then more recently um Whoop came out with a paper that um like looked at the how exercise near sleep near bedtime affects people's sleep quality. So they found and this was published in nature and this was cool because this was one of these new types of studies that are done outside of the lab and allow uh people to part participate and you sort of can validate science in a very new

17:13

way. So what we're doing is uh we're we're doing a new style of um of sleep studies and uh I'll tell you a little bit more about it how it works. So we wrote some software that takes any works with any wearable. So it works with Whoop, Aura Ring, your just phone and normalizes those scores across different devices. So this is kind of cool because now you people can compare the Whoop to the Aura to the Sleep 8 to just having a phone and you can generate a sleep score that uh should be somewhat normalized so that they can be compared. Um and then we give people crypto to encourage people to participate in the trial. We give people crypto tokens for how many hours that they sleep and if they answer

17:54

questions at the end. And the idea is they have equity in this data set. So now uh the more people the people that participate are actually incentivized in some way by linking them to the final outcome of the of the study. Um so you can track patterns, you notice patterns, you earn crypto as you sleep. Um people there's actually a lot of people in the audience who are on the app already, which is kind of cool. I see um some familiar faces who are some of the top sleepers in New York and globally, which is the first ever sleep influencers, which is kind of funny, but I think this is going to be a thing. I think you're going to have people competing on biomarkers. And what's really really cool actually is you can when you have thousands of people on the

18:34

app, you can do research a thousand in parallel. So you can do research almost a thousand times more quickly than sort of a serial more general study. it's much you know sitting and seeing the data come in it's actually quite fascinating. Um, so we have the best sleeper and we're trying to figure out what impacting sleep in like a in a in a validated model and uh giving people an incentive to participate by giving them issuing these tokens which will if someone writes a paper they can get their name on the paper they can get equity in the data set if it's sold all sorts of things. But this is sort of like what is decentralized science where you're taking using tokconomics and blockchain to incentivize the creation of

19:15

science. Um, so currently we're doing a study on like sleep masks and mouth tape and uh red light glasses and magnesium to figure out the effects that each of these things have. If it's n equals 1, if things just matter for each individual person or if there's some collective learning that can happen. Um, and then one of the more interesting things is based on how much tokens you earn, you can then use those tokens to bet on people on bet on things like do men sleep better than women? you do New York sleep better than San Francisco? And you can start to play with these tokens and incentivize people to use them for actual for having utility around them. Um, one of the things we found

19:57

already in this is that uh people's sleep score seems to decrease by 1.2 points based on how many based on the number of hours um that they ate before bed. So, this is a very preliminary finding, but I think it's cool that we're actually starting to generate data and generate findings that are that are new. Obviously, this is just the beginning, but I think it's cool to show that we are generating data and finding interesting things. Um, some of the other things is like this allows for much more nimble things like you can just look at the data and see it. You don't have to like care. It's like it's just like starting up a GitHub project. You don't you don't have to like write a research proposal and get it approved by a million committees. Science here feels much more nimble.

20:38

Um, so ultimately I think the goal is like we want to be able to figure out longevity and have markets around each person. So you could have a prediction market based on people's wearables and biomarkers to be able to predict the outcome of how long they're going to live so that you can uh accurately predict which types of interventions people should do to increase their lifespan. Um some things uh this is so this is stadium science. Um we if you're interested to download the app and participate in the first ever sleep clinical trial, you can you can uh download it here. Oh, lots of phones.

21:10

Amazing. Um if you want a job, that's the second one. We are hiring. We especially we have a great coding team now. Uh we have a great um design team. If you are like a business development type person and are very good at that, um we are looking for a very talented person in that respect as well. So, that's the second one. Um, the third one is if you're an Android user like myself, we just released a version, but it's still in the app store thing. If you want to get access to it, we can send you a code um to do that. You can follow us at Stadium Science on Stadium Science on Twitter.

21:44

Our website is stadium.science. Does anyone have any questions around decentralized science or sleep or DIY NYC? Yes. You can just shout it. It's fine. Yeah, no worries. Um, how long has the token been out for? There's no token now. It's just points, but people are acquiring the points. Um, and it's around like a month like you will or we we've had it for a month and TG I don't know I don't know when it's going to happen. Yeah.

22:17

Um, I guess one general question is offline but like I guess in the audience that how do you address the fundamental factor that all the fact that people yeah I mean I think all of it's going to be transparent so that you'll be able to see um if you if you suspect someone is biased that will be available to see whereas in a normal paper um you can't even see that at all. So any data that's extrapolated out from this data set should be looked at more carefully either by a human or by an AI. And so I think different people are going to analyze the data differently. Like two

22:58

different sets of eyes could analyze the same different same data set and come to two separate conclusions. But if that data is all available and um has provenence to it, then you can actually figure out which which type of you might be able to figure out who who's been biased and either include or exclude them from your um research. Yeah. Go ahead. Um, so who would pay for this? What was it? Who would pay for it?

23:24

What's the demand side? I think a lot of people would pay for it. So I mean one is people that want to do research, people that want to validate a model for um like sleep studies. So like sleep 8 or aura or something might be interested in purchasing this data set. Open AAI might be interested in purchasing this data set to fine-tune their model for health data. Um maybe uh people in in New York looking to figure out when people wake up so that they could you know give them coffee or something. Um all sorts of I think there's many different uh avenues. Yeah. How much? I don't know yet. That's still be determined.

24:01

Oh yeah, one more. Oh yeah, go ahead. I have a question. How do you control the placebo effect and uh how do you set up a control group with a disord? Well, I mean, it's all tracked using like a wearable. Um, so Whoop or Orura Ring and uh like that's a pretty hard thing to fake like getting how much REN you're going to get or I mean so uh I mean you there are ways to I don't think that's our first order priority, but you could have people who opt in take a placebo take a pill that they don't know and they don't know if it's magnesium or if it's a placebo. And so like two people people put in get an envelope and

24:43

they have they mix up they put in uh a pill into an envelope that is known to be a placebo and one that is known to be not a placebo and each one has a QR code on it. They scan all the ones that have the placebo. They scan all the ones that have don't have the that are the actual thing. They put those in a in a bag, shuffle it up, scan the thing that they take, take it, and that's a way to do like a selfadministered um blind study. So, you could control for that. It's not something I'm super concerned about right now, though. So, for example, like exercise and sleep, if you administer to just general public and well, they they probably have a preconceived notion of whether or not exercise before bedtime will help or

25:24

disrupt my sleep. And those sort of in fact, I guess we aren't showing this. Yeah, I mean for exercise that's a really hard one to control for uh doing a blind study. I mean there was a recent one on ketamine for depression and like the reason that didn't pass the trial is because like people were obvious on if they were taking ketamine or not. Um I'm not so concerned about that immediately but that maybe we'll address it later. I don't think it's like a top priority.

25:51

Yeah, go ahead. Do you have enough data to show whether or not daylight savings ampers or it's better? Not yet. Um because we just started this after that happened, but um I think that is cool. You could you could we you could do that study, right? So I didn't think of that study. You did, right? And that's really neat because now you can just like propose that study. We have like a Reddit, I didn't show it, but we have like a Reddit style um experiment list where people can upvote and downvote things for people to gain things and then the ones that pass a certain criteria get deployed onto the test bed.

26:27

Yeah. Could you do something with the congestion pricing? Congestion pricing. That made it. Oh. Oh, what do you mean? You have enough data for how does congestion pricing affect sleep with less cars? Oh, yeah. Exactly. So there's another example that that's the power of crowd sourcing, right? So I didn't think of that, but you could definitely do it, right? You could say we get people's people opt into giving their um broad location. It's not super precise, but we know where people are located. So we could tell if people are located in a congestion area and then do that type of study. How does congestion price pricing affect sleep?

27:02

Uh yeah, can you speak a little bit about like a few sentences about the privacy? Can I share this with my friends in Europe? Yeah, definitely. So, we use uh Nillian compute which is like an MPC MCP compute where it's encrypted on the device and then sent and then stored and then the the data is stored the data is computed over the back of that which is which is so it keeps it private. Um there's also a lot of features in there to make your data public or private depending on what you want. Yeah, people in the back. Um go ahead. Okay, any more?

27:36

Three, two, one. Okay, without further ado, guys, people in the back, we can actually hear. Um, and we're going to take it over to Alex who I met as a we were at Stanford together and we studied uh computer science or he's I studied computer science and he studied um bio mathematics bioinformatics sorry and um very interesting guy. He has uh an intense background in both uh research and in medicine. So, um, we are very proud to have him. Give him a round of [Applause]

28:17

applause. Let's go. Let everyone smile. Three. One more. Focusing. Okay. Sweet. Okay. Okay. Excellent. Well, listen, thank you so much, Mikey. So, Mikey and I uh lived across the street from each other when we were in graduate school. Uh and it's such a big pleasure to uh to reconnect,

28:57

especially lived across the street from each other in graduate school uh in California. It's a big pleasure to be here to uh to reconnect. Um and so today I'm going to be talking about biomarkers for longevity and I come here knowing that Brian Johnson has recently given a talk uh here at DAI. And so I'll just preface by saying I'm no Brian Johnson but I hope nonetheless uh you love the talk and uh you get a lot out of it. Um because I'm a physician uh I have to give you disclosures otherwise I'll lose my license. So I have some equity in a biomarker company called liquid cell from my PhD thesis and I have a venture capital firm called September. So I want to start the talk by talking about age and when I was

29:37

researching for this talk I came across kind of a striking statistic which is that age is probably more than any single thing the greatest marker and predictor of mortality at any given time. I thought that was very striking. And so age is very salient in two periods of your life. The first is on the day that you're born. Infant mortality, even though it's 2025, remains a huge source of mortality, globally, developing world, um developed world, across everything. Um so you have a high chance of dying on the day that you're born. But if you make it, you end up in kind of this valley of safety, but then age creeps up on you again. And as you get older, your mortality on any given day is actually really governed by

30:18

age. Uh and age is very uh strongly correlated with that mortality. This is data from Sweden where they kept records in the 1800s. And remarkably, as you might expect, this curve of mortality versus age goes southward. It drips down over time as we improved, you know, public health, sanitation, as doctors started washing their hands, some still don't. And as uh we kind of developed medical technology over time, I thought that was very striking. At the same time, we all know that life expectancy has increased. And in kind of the modern industrial era, it's increased in a sigmoidal fashion. I want to just give some further context by going all the way back to ancient times. Um, and we kind of had this linear increase up into

30:59

the the medieval plateau, I'll call it, uh, and then into our industrial sigmoid function. But how did this sigmoidal climb happen? Well, it turns out around the inflection point, we had things like public health, we had infrastructure, we had vaccination, we had sanitation. And that gave rise to this uh this increased slope uh in the industrial era in the developed world. We then had kind of in the early 1900s post World War II or around World War II, we had modern antibiotics. We had penicellin and then further we had kind of these modern therapeutics that further boosted um longevity. Then kind of in the 70s uh and and then you know a bit further we started to tackle chronic diseases. We got better at diabetes. Um we got better

31:40

at tackling heart disease. we got better at cancers. You know, breast cancer is one that really uh comes to mind. Especially me as a cancer doctor, I'm very excited about and we've done a tremendous job doing that. But here we are in 2025 and there's something very sad that's happened which is that there's a stagnation. Um and this is you know kind of as a physician heart disease have cancer.

32:02

Oh sorry there's a bit of feedback. Okay. So there's a little bit of stagnation uh in in 2025. Um uh and so uh uh uh things are really stagnated. Um and then kind of like with everything, America really potentiates global trends. And so in America in particular, life expectancy has really plateaued and has even decreased. You might notice that there's a blip there. You might imagine that's from the COVID pandemic.

32:28

But even without the COVID pandemic, um life expectancy has really started to plateau in the US especially. What's even more concerning is health span. And so those of you who were at Brian uh Johnson's talk will probably know health span is this measure that takes into account lifespan and also the years of life live free from disease from disability and from suffering and is a key key metric. The problem is that health span has not only not really increased it's actually started to contract especially in America. America has an unfortunate distinction of leading the world in the gap between life expectancy and health span. Meaning that health span is really lagging life expectancy particularly in the US and

33:08

it's a bit small on the slide but the United States is highlighted as kind of the top point in this graph unfortunately from this wonderful JAMAMA paper that I found. Um and then the other unfortunate thing is that the graph on the right shows that health span lifespan gap in the United States is increasing over time. So health span in America is contracting. So I want to ask you how should we think about the measurement of health? Well to do that I want to add some context and go back in time. Let's go all the way back to 400 BC where life expectancy in ancient Greece was about 25 years. Live fast and die young. And you know Hypocrates was writing his apherisms. Um and so apherism number 44 just happens to be you know um people

33:49

who might be a little bit overweight. they tend to not do as well in terms of longevity than other people. And so kind of this is one of the first times that somebody talked about a marker uh of longevity that I could really find. And it turns out you can tell a lot about health before you even do invasive testing like blood tests. Um this is a sculpture by my here depicting an Olympian. And even Hypocrates himself would know that this Olympian has a lot of life left to live compared to somebody who's pale, who's kind of in bed and kind of sick and bedridden. Um, you know, you can even just tell from the vibe of these two pictures. Oh, yeah. Like this. Yeah. You can even tell from these two pictures. Uh, there's a huge delta between them. In the medical community, we have a term, especially

34:30

when a physician doesn't want to examine a patient. I'm not endorsing non-examining a patient, but a doctor might say, "Looks good from the door." Uh, LGFD. They might write that in their note. And certainly the the picture on the left looks much better from the door than the picture on the right. And you kind of just get this vibe. So um but how can we really quantify this vibe about lifespan? So starting in the Renaissance era uh there's this guy uh Santoro Santoro and he wrote a book called Dstatistica Metachina. Um in it he was kind of like a Renaissance era Brian Johnson. Uh he was weighing himself every single day. He was tracking all the food that he was eating. he was measuring like how much he persspired an insensible loss and

35:12

he's trying to correlate these against his healthcare uh outcomes and so I thought that was very fascinating. Then there's this guy called Quint uh who actually invented the body mass index um and then Ancel Keys renamed it to its current name of of of BMI uh in the 70s and this is another example of an early biioarker that was made finally in 1969. This is one of the seminal papers or first uh uh uh uh times that somebody wanted to make a measure of how fast a human would age.

35:42

And they picked up on many different metrics. Some of them we know very well today like hand grip strength which we continue to use. Um blood pressures which are certainly very validated. We use cholesterol and then something called the vibrometer. I can't imagine what it does. Maybe it measures vibes. I'm not sure. But vibes visible and invisible. We quantify them. this gut feeling that we have when we see a patient as a physician or when you you feel something yourself, we quantify them using a biioarker. And so what is a biomarker? A biomarker is a measurable indicator of a biological state. They help us to understand and measure age, health and longevity itself. Now a little bit about me. How did I come to study biomarkers? Well, when I was doing

36:22

my PhD, I became very excited about bionirmatics and about cancer. There was a tremendous amount of activity happening in cancer during the time of my graduate studies. Around that time, amunotherapy was really, you know, bursting onto the scene and very popularized. Amunotherapy for skin cancer won the Nobel Prize in 2018. Um, and it turns out that people who have metastatic melanoma, metastatic skin cancer. Um, about 50% of them will have a dramatic response with imunotherapy whereas in the past it would have been a death sentence. But it's a double-edged sword because amunotherapy seeks to unleash the immune system against the cancer. But it can also mediate an autoimmune attack against your own body, sending about half of patients to hospital and even in some cases it can

37:03

cause death. So in my graduate studies, we profiled the blood of patients before they got amunotherapy and we found cell phenotypes or types of cells that are very abundant in patients and we figured out which of the patients based on their cell phenotype makeup from this blood test would go on to have a severe immune related adverse event. Um and so this is our AUC from the validation cohort. And you know it was a big pleasure to do this and I continued my interest in cancer. I finished um medical school and then uh I'm actually just finishing my internal medicine residency uh imminently and starting my oncology fellowship um any second now. Um back to the the gerine topic of this talk is age. Um so part of aging is this genetic

37:45

program of development and another is the accumulative of substances experiences that we're exposed to during our lives. I want to ask you, can you guess somebody's age just by looking at them? And can we fit to age using parameters that we find in a blood test? This next slide demonstrates a very important paradigm in the field of longevity science. Um, so what a lot of researchers in this field do for the past, you know, 15 years or so is they take all the patients chronological ages in a cohort. We call it a training cohort. Then they do something like a blood test. They have multiparameter blood tests with tons of data and information and they figure out, you know, for each given age of a patient, what are the parameters in that blood

38:26

test? They then fit a model. In the old days, it was statistical learning. You might have run it in R and Python. In these new days, you can run an LLM. You can do deep learning on it. And you're kind of fitting a model um to figure out how you can predict age from a blood test and from biomarkers. You then go ahead and you plot chronological age.

38:44

The age that shows on your passport or your driver's license against a biioarker predicted age. Then the next thing and this is the key step is that we create a biioarker age gap. And I want to really hammer this home because the rest of the talk is really predicated on this and it's a fundamental concept in longevity science. We calculate a delta between the chronological age and the biomarker age. And so it turns out that if you have a positive biomarker age gap, you're going to be predicted to be younger um than you actually are. And if you have this negative biomarker age gap, you might be predicted to be older than you actually are. And when a relative or somebody you know is telling you, oh, I'm actually, you know, my blood is younger than my actual age and

39:24

they've done a blood test from an insurance company or something like that, this is what they're referencing, more or less. So let's go into our first academic paper. So this is a really wonderful nature medicine paper that came out uh two years ago now. And in this one they uh they profiled the blood of tons of patients in the UK bio bank data set um thousands of patients and they had an ensemble of uh proteins that they did feature selection on and they fit a machine learning or AI model on to predict age using proteins. In this their biomarker is called protein age pro age and their biomarker age gap is called the protein age gap. The plot on the left is the UK bio bank showing their protein age plotted against um age

40:05

in the in the cohort. Um and they show a very tight fit. The piercing correlation is 0.94. And this is their test set. But the thing that's remarkable is that although this was trained in the UK, they profiled the blood of a Chinese data set and they got an extremely tight fit in a a kind of a blinded test set and a piercing correlation coefficient as well of 0.92. And so even geographically, culturally, you know, genetic or or e ethnically different groups, they actually were able to fit age uh in kind of a blinded unbiased way in a training and test cohort. And so what's the big deal? What can you do with this protein age? Well, it turns out that if you kind of have a very unfavorable protein age gap, um it turns

40:45

out that you rate your own health very poorly. Um you have a slower walking pace, um you have a worse self-rated facial aging. Um and then on the other hand, if you have a very favorable protein age gap, your hand grip strength is very strong. You have better lung function. Um and uh you actually have better fluid intelligence. Um so I found that that to be quite fascinating. But where do we go from here? So we can actually measure just really a ton of things. And so we can measure, you know, physical characteristics like weight, height, BMI, etc. We can measure things with things like our Apple Watch and our wearables. We can do MRIs on the brain.

41:22

We can do things like V2 max and heart rate variability which some of you are probably measuring right now as we speak. Um, we can also look at molecular markers and that's kind of the area that I focused on in my PhD. You can look at protein like the paper I just showed you. You can look at the transcriptto like mRNA sequencing. You can look at epigenetics like methylation data and you can look at the metabolism as well.

41:44

I want to next show you a paper from the epigenetics uh area. Um this is one of the seminal papers in longevity and anyone in longevity who's an expert in this room probably has read this. This is kind of the Horvath biological aging clock. Um and this is a 2013 paper. Um and here the biioarker was methylation data. So the epiggenome how do your genetics change over time through you know nurture rather than kind of nature.

42:08

Um and their biomarker age is called the methylation age. With that exact same paradigm this like fitting the age and the age gap. Here on the left we have the training set of bidding age and on the right we have their testing set and they actually had excellent results in a held out blinded testing set to predict age um based on their machine learning model. One of the great things about this is that you can if you happen to have your methylation data um or you're able to get it, you can upload your methylation data to their website. Uh screenshot is here and you can figure out your own methylation age. And I thought that was totally fascinating. So if you have a genetic sequencer lying around, you can you can go ahead and do this. Um, another thing that I loved about this paper and I thought it was very provocative is that you can

42:49

actually probe methylation age from each tissue in each specific organ. I think this is very clever. So you can probe the methylation age of your liver, of your lung, of your kidney, um, of your brain. I will note some of the tissue sites are a bit underpowered and the muscle one doesn't really work. But I think that age per organ system is very provocative and kind of generated a lot of thoughts in my own mind. And then you know what's the big deal? Well, it turns out that people have continued to study this uh this methylation clock and have benchmarked it against other methylation measures. And this is a JAMA article that you know came out in the last 2 years and it turns out that epigenetic age certainly does in independent

43:30

researchers, independent cohorts uh correlate with longevity, mobility and cognitive function. So, you know, there's this hypothesis that this stuff actually really matters and really works has real world uh uh bearings and outcomes. So, another question I want to pose is are there multiple dimensions to age? And so, I made this little schematic here, but I kind of started to read these papers and then think, you know, might there be an age for your cardiovascular system? And I'll plot it on one axis here. For those of you taking linear algebra, this will kind of ring true. Um, and then there might be like a metabolic age that you might have for your metabolic system. There might be a neurological age for your nervous system as well. And then I'm kind of constrained by a two-dimensional slide

44:11

here, but there might even be more axes or more interactions like with genetics. There might be interactions with the environment as well. And then you might imagine that the diseases that you end up happening uh having uh are kind of like a projection upon these axes. It's the makeup of all the ages and all the you know vulnerabilities and all of your major organ systems and their interactions with the environment and your genetics. So you know heart disease might be a component of you know cardiovascular metabolic and then a tiny bit of neurological as well. You might have emphyma or COPD be a component of these things. You might have cancer be a component of them. Autoimmune disease nerve degeneration and stroke just to name a few examples. I want to just loop

44:52

back just very quickly to the protein age paper because in their supplementary material there's a very provocative uh figure and what they found is that for protein age they're actually able to stratify patients who had kind of low protein age gap so a very favorable age gap in green um all the way to poor protein age gap unfavorable in blue and it turns out if you have a very poor protein age meaning that the protein age uh uh that by your blood is much worse than your chronological age, you're much more likely to have a chronic disease.

45:23

And this can be autoimmune like rheumatoid arthritis, it can be, uh, you know, a type of cancer, it can be heart disease, it could even be a stroke. And I found that to be totally fascinating. So, uh, another question I have is, do biomarkers generally hold across ethnic, cultural, and geographic differences? And we saw in the protein age paper that they trained in the UK and they validated in China. This other paper um, said kind of similar things.

45:46

They looked in Canada, Korea, and Eastern Europe, and they found kind of the same thing that things really do work. We have a lot more in common than we have different, but they did do some fine-tuning across all the populations. And so I thought that was interesting as well. Um, so I actually did the final internal medicine call shift of my life overnight on I think it was like Saturday night. It's all really a blur, but I I kind of slept at, you know, 10:00 a.m. and I woke up at a time and I'm ashamed to say I was on on my phone in my bed. Uh I don't want to admit it, but I opened Twitter and because I was preparing for this talk, I saw kind of a very apropo tweet from Chumath Polyhabitatia. And in that tweet, he kind of goes into discuss how, you know, he's been tracking his V2 max with his

46:26

Whoop, his heart rate variability. Um and these metrics have kind of gotten worse over time for him apparently. But at the same time, although his metrics have gotten worse, he's had an amazing life the past few months. He's drank incredible wine. He's spent time with his wife, with his children. he's played poker with his friends. And so, you know, it was kind of interesting, you know, what's the point of all these markers? Um, and you know, do they really matter to the average consumer?

46:51

So, Kevin here weighed in and actually had an interesting opinion. Uh, Kevin says that body fat, LDL, like your cholesterol and blood pressure are much better validated than HRV, heart rate variability, and some of these other metrics. Um, Brian Johnson, the goat, of course, chimed in here, uh, and said to Chimoth, "Let's just focus on a few key things and do one thing at a I think very sage advice from Brian Johnson here. Um, Eric says, "I actually don't trust this stuff at all." Um, Chadon says, "I extol the virtues of lifting."

47:19

Very reasonable. Uh, Daniel says, "We're measuring way too much stuff." Reasonable as well. But Will chimed in as well, and Will actually happens to be the the CEO of Whoop. Uh and this is actually very fortuitous and very fortunate for me because um I happen to be giving this talk and will decided to release Whoop's health span feature on May 8th just in time uh for Dside. Uh and so Whoop's health span feature gives you kind of your estimated age. Um and then it gives you this delta age as well, how much younger you are uh than your chronological age based on all the metrics that Whoop uh that Whoop kind of aggregates. And so I don't have all the specifics on how it works underneath the hood. Um, but it does aggregate these

48:00

things. It puts them into a machine learning model. It'll aggregate like your age, your weight, V2 max, how much you're exercising, and and all those sorts of things and kind of give you this this Whoop uh biological age. And for those of you who have the latest edition of Whoop, uh you can actually find this and you can you can you can figure out kind of these metrics and and look further into your health span. So, I think that's totally fascinating. As you know, there's tons of consumer uh consumer wearables that are out here. Um, I won't go too deep into them and I'll keep it more academic. And then, um, hearkening back to Brian Johnson, you can even, uh, buy Johnson's, uh, blueprint, uh, uh, you know, system. And you can you can measure your age, uh, per his company with some blood tests that that they offer as well. So, there's a ton really

48:41

on the consumer side, but I want to bring it back to biomarkers and AI uh, just to kind of round things off. So I wonder how will AI and large language models act as biomarkers and how will it allow us to understand and even uh extend longevity. Um the first paper I want to just quickly touch upon and I think this uh this field is a little bit immature but that uh researchers are actually looking at MRI images with deep learning to scan your brain and kind of giving give you an age on the basis of your brain. Um so there's a lot of activity in this field. It turns out that your brain age is very correlated with mortality. Um, but there's kind of mixed data so far on if it's correlated with your cognitive function. Um, so I think

49:21

probably a lot of potential there and I'm going to be watching that very closely, but uh very early for that. This is one of my favorite papers that I came across. So this uh again came out just now. It's just breaking. It's it's May 8 and this is from a bunch of cancer doctors and this is very near and dear to my heart. um these scientists here and these doctors they actually trained on a data set of publicly available face data to predict how old you are on the basis of what your face looks like and these are just healthy patients. Then what they did is they had cancer patients from the Netherlands um and from Boston. They had lung cancer um they had other types of cancer and they applied their facial recognition or facial age u uh uh AI algorithm to the

50:03

cancer patients to figure out you know a face age of them. The thing that's striking is that uh facial age actually can predict overall survival. And for those of you who are familiar with Kaplan Myer curves, it turns out that the older uh your face age score is um the less long you have to live unfortunately. Um but the interesting thing is that uh it can really stratify people and and significantly so they achieve statistical significance with just this AI algorithm. It turns out that your facial age in a cancer patient is more predictive of survival uh than actually your your true chronological age. And I thought that was uh extremely fascinating. And that's kind of borne out in this forest plot on the right.

50:46

One thing that was interesting as a side uh is that it it can actually predict who's a smoker and who's not. It turns out that your facial age uh AI score is really correlated with your smoking status and is significantly so. Um and then as a physician myself, I you know I was you know a bit sad to see this but also really not surprised at all. Um it turns out that uh facial age is actually incredibly good at figuring out um uh your survival at six months uh from when it does the prediction. The area under the curve or kind of an accuracy score is 0.8 for 6 month survival predicted by the facial age AI algorithm. For doctors on their own without any assistance, it turns out that it's much worse. It's

51:26

0.6. So doctors almost it's like flipping a coin to figure out who's going to survive at 6 months. For doctors who have the patients physical chart, the electronic health record with them, um it's much better. They get a lot more data. And then doctors who have the facial age score and um have their own clinical judgment and the chart, uh they perform as well as the facial age score. But we've seen this time and time again where actually an AI algorithm trained very well and precisely is very good at predicting clinical outcomes.

51:54

And unfortunately physicians with algorithms together, they're actually not often able to beat an AI algorithm at predicting outcomes even when doctors have assisted AI at their disposal. Um so you know super curious, maybe not the best news um for doctors like me in my shoes, but uh as a scientist totally fascinating. I want to draw your attention to kind of one final paper and this is more in the uh nebulous like uh cutting edge part and this one actually is only in bioarchchive. It's not in a peerreview journal just yet. So everything with a grain of salt but um in this one the authors uh read the electronic medical record um and uh uh well they used AI to read the electronic medical record and

52:35

then predict age from reading a patient's um a patient's chart. all the notes that doctors, nurses, dieticians, PT, OT, everyone writes about a patient. They had the AI read the notes and figure out the age. Um the correlations are pretty good like the Pearsons are, you know, 75 and so and and very significant. But I think the thing that was the most striking is that um the AI is really able to very strongly risk stratify patients and predict overall survival. And this is another very striking significant Kaplan Meyer curve showing how uh just the AI trained on the electronic medical record is able to stratify patients. The authors then went on to make an omic age which is really multimodality and it's AI trained on proteins, metabolites and clinical variables altogether. And with that omic

53:16

age they were able to do a lot of cool things. One is that they're able to predict 10ear survival very well and it's kind of exemplified in this uh uh like uh AU plot over here an ROC plot uh kind of in yellow and it beats out um the just the chronological age here in black. um the EMR age actually uh so the electronic medical record age by AI does very very well impressively and then for the major chronic diseases to predict who's going to have a new onset disease so an incident disease um it does very well especially you know for things like diabetes um one thing that's beyond the scope of my talk uh but I thought was interesting is that open AI actually is also dipping their toes into longevity science I

53:58

don't want to take up too much of your time so I won't go into that but um I think you know it's just such a hot topic and there's so much exciting work that's being done. So, you know, coming towards the end, what are really the next steps here? And I want to just, you know, as a scientist say just because something correlates with age doesn't mean that it actually causes age. And there so, you know, we don't want to get into a correlation causation fallacy.

54:18

But at the same time, I don't want to lose hope. And if there is real biology at play and if that can be borne out and real experiments um biomarkers for age may actually give us hints at new therapeutic targets to actually target the processes of aging. Um and so I think that's totally uh uh very provocative. Uh there's also a whole new set of startups being created in the longevity space. I'll just say the scope of this talk is really focused on academic research, but in the private sector there's really a tremendous amount of rigorous activity and many of the papers I showed you actually have a startup company that's accompanying them and you can even use some of their products today. One question I have is how much can we actually personalize recommendations using this biomarker

54:59

data and can we actually beat what all of our mothers told us which is just you know go outside and run eat you know vegetables and don't smoke and one of the ways I think that biomarkers can help us personalize uh uh recommendations from a physician or from any healthcare provider is by having this multimodal approach where we can have an age per organ system to really weigh in and and how can we predict the onset of a very specific disease so we can start screening uh for it earlier like COPD or like diabetes based on early warning check engine light signs in a specific you know domain like cardiovascular for example. Another uh interesting thing is what can we do with all this data and a huge need uh is for clinical trial

55:39

endpoints. So there have been clinical trials which are actually using methylation age and some of these other age metrics as endpoints. One is a calorie restriction clinical trial. One is a trial for the drug rapomy and anyone who's read the book outlive by pureia um he goes into rapomy a great deal in the introduction of the book um and another is for plasma infusion and so all these clinical trials registered with the NIH use uh these biomarker ages as an endpoint one thing I really want to hammer home is that there's a key problem in clinical trials for most diseases right now in that it's actually very challenging to use mortality or lifespan as an end point because it's often very many decades away people with diabetes often live for, you know, 10,

56:19

20, 30 years longer, but they're living with the side effects and and, you know, the crippling aspects of diabetes. So, for chronic diseases, we really need to have a way to measure um the impacts on lifespan and health span immediately. I think that biomarkers are the instrumental tool that allow us to have a readout of increases in health span induced by the intervention that we're testing um uh in any given clinical trial. And so, they're absolutely key to this coming revolution.

56:44

I'll just end by going back to Chimath here and you know he's asking more philosophically what am I supposed to be getting out of all these biomarkers and what's the use of everything? Is this all just really a vibe and is how you feel matter? Um is what you look like from the door when you're examining a patient is that more important than all of the biomarkers? Um I well I think that there's a lot of utility still to measure things and um just to come full circle I want to talk about where we came from. So, one of the great challenges of our time is that human lifespan has really stopped increasing.

57:15

Uh, this is devastating. We've stopped living longer as a population average. Uh, furthermore, health span has actually contracted and I'll just hearken back to this plot of the health span lifespan gap increasing in America, meaning our health span has contracted. But within this data lies opportunity. Uh, there are key areas where health span is very low and the gap is greatest. And these were driven and I know the text is very small by like things like mental health, substance abuse disorders, um respiratory disease like emphyma, neuro degeneration like Alzheimer's and cardiovascular disease.

57:48

These represent key challenges but also key opportunities for therapeutics and opportunities where we can measure the outcomes of therapeutics instantaneously with biomarkers. I want to go back to this plot about stagnation. And so, you know, the increase in lifespan from ancient times was really the product of human innovation in therapeutics, public health, infrastructure, sanitation, and even diplomatic efforts for peace. We stand now in 2025 at a time of life expectancy stagnation and health span contraction. And it's incumbent upon us, all of us in this room here as innovators, scientists, doctors, and patients alike to accept the baton passed to us by our ancestors and increase health span for our own

58:29

generation and for generations to come. I think in the dawn of the LLM era, we might be able to enter out of a age of stagnation into an age of health abundance using the new tools at our disposal. And there's a tremendous amount of opportunity for us to make a real difference in this urgent need area. So biomarkers will allow us to not only measure our own progress in our own health journey with consumables and wearables but also the progress in the critical human endeavor of increasing health and lifespan uh all together.

58:59

Thank you so [Applause] much Q&A. Does anyone have questions? All right. Yeah. Go ahead. Are you familiar with like the dune and face test? Wait, there's people who are the mic and then Oh, they they they speak with the mic. Yeah. Oh, sorry. Okay. And then you can pass it to them. Yeah. Okay. Are you familiar with the uh doing it in past? Yeah, I I I saw I didn't I didn't uh Yeah. Oh, and I take it back. Okay. Thanks. Sorry, I'm

59:40

new with this. So, repeat the question. Yeah. Am I familiar with the June base test? Yeah. Yeah. I I I define it. Can you find what it is? Yeah. I So, uh, no, no, really hold it. Oh, sorry, guys. Okay. Okay. Excellent. Uh, yeah. No. Um, I saw it preparing for this talk, but I didn't go into a ton of detail, so I won't kind of speak out of my depth on that one, but I know it's one of the very promising, you know, tests for health span and longevity as well. Yeah.

1:00:07

Okay. Yeah. What's the test? The DUN base test. It's another, you know, competing test. Um, I forget which modality it interrogates, but you know, another promising exciting test for health expansion. They have a leaderboard and I think Brian Johnson's number one, but you have to do the test like sequentially like three times over six months to be number one on the leaderboard or like top 10. Oh wow.

1:00:28

Okay. Yeah. So, for those who didn't hear, they have a leaderboard and you can you can figure out, you know, who's the top one and Brian Johnson's the top on the leaderboard. Um, so I think that's quite fascinating, I guess. Yes. I'll give the mic to someone who hasn't uh spoken yet. Anyone you guys have? None of you guys just spoke. Can you pass it to the lady just over your right shoulder?

1:00:47

Hi. Go trees. I went to Stanford. I don't know how I did not come across you when I was living in Eskanito Village. Okay, if you're also living there, but um I have a question. Do you know if any biomarkers seem relevant for health disparities? My geriatrician physician friend is asking. Yeah, that um I guess I'll grab contact with you. You know, it's such a good question and I I didn't I didn't uh pick up on a specific one, but we do know that uh like there's methylation uh outcomes from living in poverty. Like poverty

1:01:28

actually changes your methylone. Uh you know, the toxicities in the air, the stress of being in poverty. Um all these certain things kind of weigh upon us in our genetics. I didn't come across on a specific biioarker for health disparity, but I can imagine if you go back and reanalyze some of these cohorts and some of the data. Um, you will be able to find something. I'm I'm sure you will.

1:01:48

Uh, you know, in geriatrics, uh, I, you know, I've done a geriatrics rotation and I work, you know, primarily with adults probably over 70 to be honest. The median age of my patients uh, is probably 65. And it breaks my heart to see, you know, patients from impoverished backgrounds. And so I think that they deserve access to these health span biomarkers all the same. And so I think a key challenge for us is to be able to distribute uh this good to to all sorts of populations. So I don't have an exact answer for you but I might you might imagine that reanalyzing some of this data um stratified by you know income level there I think there will be signal there. I I I really do.

1:02:25

Thank you. So great talk but I I think the question that I have and it's more of comment isn't this whole idea behind biomarkers and end points only dependent on where the science is currently. So my understanding you're an oncologist I'm an oncologist and we know that even endpoints like PFS when it comes to things like immune therapy don't do as well and nobody would dispute that immune therapy isn't beneficial. So like as science advances, isn't it feasible to say that the current biomarkers and what they show and the current endpoints that we use easily going to get

1:03:07

replaced? So like do we and I know a lot of private companies are putting a lot of money and stock into some of these things by saying well this is absolutely first science. It's hard science today till the science actually changes and then it's yesterday. Yeah, I I I think that's an excellent comment and um you're an oncologist with a lot more experience than than than I have. But I think it really rings true and I think part of biomarkers is uh you know we're looking under the lampost and we're examining things that we can actually measure and so we have this paradigm of you know biomarker age uh but this may not be the paradigm actually forever um and so that might change um and then you know there's a huge financial incentive to show that an

1:03:47

endpoint changes that the p value is different or that the box plot of group A is different than the box plot of group B um when you stratify by a key biomarker and I agree like the goalpost may very well change over time. And I think one of the key most critical questions is are these biomarkers going to actually translate into benefits in qualities quality of life years um into real health span however you want to measure it um by having a better longer you know uh more symptom free life.

1:04:15

Those are the uh major questions and some of those questions require years to kind of really bear out in the data and I do think that the goalpost will will move over time as we become you know more uh expert in this field as AI is better and we have more access to data but uh I I think it's a great comment and yeah thanks for thanks for the comment I appreciate it.

1:04:39

Does that which does not kill you make you stronger? Um Kanye is canceled is the first thing I'll say. Uh so uh shouldn't have even said his name and but uh what I would say is uh uh I mean I I really I really think so. Uh I think it's very disease dependent. Uh and so it turns out that if you have damage to a certain organ like to your lungs um or to your heart and things like that um you do get weaker each time you come back into the hospital. And so I actually think you can probably capture that with certain biomarkers. Um certainly we can see it, you know, in our heart failure patients with their um their BNP marker over time. It's kind of

1:05:19

climbing and uh you know, other markers like that. Um but I, you know, I I like to think maybe I'm just young and idealistic that, you know, psychologically or mentally we become more robust over time. But uh yeah, I I think very provocative statement. Thank you. And look at you next. just don't bump the camera. Uh, great talk. Thanks for that. US has way worse health outcomes despite way higher spend than most other western countries.

1:05:47

What would you say are the main cause for that? And what would you suggest on a policy level what to do to improve that situation? Thank Thank you so much for the question. I think that's uh an incredibly uh enormous question and and a fantastic question. Um, I'm actually I was born in Canada and I have dual citizenship of Spain in Canada. So, uh, I I but I did go to school in America, which where I met Mikey. Um, so I I think the thing I'll say is that um, oh gosh, it's such an important question and I'm not a healthcare economist. Uh, I would let me say it from a doctor's point of view. I think in doctors, nurses, dietitians, PTs and OT's, there

1:06:28

are so many people, the patients themselves, patient advocates, so many good people out there who want to make a real difference for patients. Um, want to do a lot of real good uh and help the people in front of them. Um, I think that things have gotten more stressful. Billing has gotten very con convoluted and complicated. Um, I you know uh there's many models across the world that uh you know work more efficiently and work better. I think anything that we can do to uh reduce overhead, reduce administration, um reduce the complexity of billing, reduce the complexity of seeing patients um on the healthcare uh provider point of view will reduce our burnout um and make us um uh you know more adapted uh fostering good

1:07:09

healthcare outcomes. But I think it's an enormous question and I thank you for it. I probably am not equipped from an economist point of view just today but uh I yeah I really appreciate the question. It's it's a huge question for America. So my question is um in regards to blue zones um has there been any thought to studying the biomarkers of people that live in these areas in terms of longevity and what are your thoughts on that?

1:07:42

So you know I think it's a fantastic question. um my you know my my ancestors are all from the the Mediterranean and I tell all of my patients it doesn't even matter what you have if you kind of walk in and you see me and we spend enough time together I'm going to tell you the Mediterranean diet is what you should be on no matter what and so uh I think in terms of the blue zones I actually would like to study you know these populations um and I I'd like to know what their their methylme looks like what the proteins look like in their blood I think that would be very curious um something I'll say is that not everybody you know in Spain or in Japan um is a paragon of health and they're still smoking and things like that as well.

1:08:19

But I would be very curious what the blood uh looks like in those areas and what the biomarkers look like and actually is there an interaction with culture as well. Is there something about family about community about having a siesta once a day uh that really you know potentiates healthcare outcomes? I'd love to know the answer to that. I don't have an answer today but I'm you know very curious about that too.

1:08:45

Uh thanks for the presentation. There are some prominent people making some statements about artificial intelligence effect of longevity. You probably heard that Dar Amaday of anthropic made a statement that AI can expand life expectancy uh double it in the next 12 years so we can live 150 years. Uh how do we get there and can we get there in your opinion? Hey, so did you say 12 years?

1:09:17

Yeah. 12 years. 12 years by 2037. It's 150 years, you know. Uh I mean, this is being reported, so I'm just never going to say never so I don't look dumb in 12 years from now. Um I think there's going to be a huge marathon uh to get there. It's going to take a tremendous amount of uh of work. there's some kind of genetic uh barriers that we're we're brushing up against. There's also, you know, cultural and physiological barriers. Um I guess one way I'll answer your question is, you know, for whom? And so I think that uh people in Brian Johnson's shoes and, you know, people who are living in um very affluent countries who have access to nutritious

1:09:58

food, who can work out every day, who can sleep, you know, excellently um and kind of have, you know, genetics that are robust enough to have it. I think that they will see uh dramatic increases in their lifespan. Um I and their health span too. So I I I really am a believer that people will start to live longer. I think the record I was reading out live by Peter Tia and I think the record is like in the like it's like 122. I might be wrong but it's it's something like that. I think that we will exceed that record um very soon. Um but I think that the benefits will be acred to you know the very uh most elite echelon of people. But then going back to our disparities question, you know, I think

1:10:39

um we have to think about how can those um benefits um and the things that we learn by doing blueprints and by being as healthy as we can, how can we distribute that to as many people as possible. So I don't know if we will double in 12 years. I mean I think that would be a fantastic outcome and I think if we do you know this goes beyond my scope of being a scientist. I hope that people uh are fulfilled and happy. Um and that's something that a doctor can't provide. I think that comes from within, but you know, what's the point of having a longer life if it's not worth living?

1:11:06

So, um, but yeah, I think that's very provocative statement and very energizing to hear. So, you know, fingers crossed. Um, I'm really fascinated with the longevity scope um, and it's particularly anti-aging. I would love to hear your thoughts on um, stem cells. I read a really interesting study on menstrual stem cells. So, I would love to hear if you have any thoughts on how that how that factors into longevity and where it may be going in the future.

1:11:39

Thank you so much. I in my graduate school, I actually worked in the low-key stem cell building. I don't know if you know, yeah, I worked in the stem cell building. Um, but that said, I had no uh personal exposure to stem cells. So, I I don't want to talk too much out of my depth here. Um the thing I'll say about stem cells is that you know remember like in the early 2000s we were talking about stem cells and you know people living longer and being superhuman and all of these things. Um but there are some real uses for them. One of the labs across from my PhD lab uh was using stem cells for osteoarthritis. So you know that wear and tear type of joint pain that you might get in your knee where you get knee replacements. And and they actually found that stem cells you know were working well there. I I I'm just

1:12:19

blanking on the exact type of stem cell and what kind of like pre-engineering they did. Um there's also been a study uh that came out in Nature. I think it was about two weeks ago where they uh tried to use stem cells for Parkinson's and there's no like it was like really phase one and like there's no they don't know if it works or not. They just know that the patients like didn't die and it's like it's quote unquote pro probably safe. Um but uh time will tell if it actually you know abrogates like the development of Parkinson's over time. Um and so uh there are people you know making real headwinds here. Um but also caution saying that there are many clinics uh in not in the U US but in other countries where the regulation is

1:13:00

not as tight um which are selling stem cells to people um for no good reason and people are spending a lot of money on them and there's actually been some very high-profile cases of contaminated uh vials of stem cells injecting into knees um and giving people septic knees and and people getting abscesses and things like that. And so there will be people who are trying to capitalize on this beautiful moment that we have where we all come together and use science for good. There will be the people who try to to capitalize that um without you know rigorous scientific practices. And so I would just exercise a lot of caution but also be optimistic is how I would answer the question. Thank you so much. They're legalizing stem cells in Florida just amazing. Go

1:13:43

Florida. Do you have a question? Hi, thank you for your talk. I probably have two questions. First may comment there all biological ages. We are actually have huge variety and considered not reliable by some and by me as well. And second one about mutilation. when we say when we look at mutilation DNA we say well DNA is not our destiny so what do you know about what is measured uh in probably I don't know markers biomarkers and if there is any generally indication of better

1:14:26

biomarkers like metabolomics or anything in this um direction if it can predict better like as of today mutilation not just by genes Um, could you two in the back be quiet? We can hear you up here. Yeah, to your to your your first comment. I totally agree. I mean, I think age is is very diverse and it can be diverse within somebody and you can, you know, you can kind of be more youthful and younger and older all within, you know, the span of your own life and it's not necessarily monotonically uh increasing. Uh I think to your second point um in terms of methylation yes like I think that um metabolomics there was a paper that I did not include about metabolomics um

1:15:06

because I just didn't have time um but there are many groups working on the metabolism as well the metabolites in your blood um uh there is a lot of cycle in there as well I think the thing that you know maybe a limitation of my talk is I don't have like a bar plot of head-to-head every single modality and which one produces a better um age and so but I I think a little bit for good reason because I don't think that there's one modality which is really the clear winner just yet and it has more to do with the machine learning the the kind of authors who wrote the paper um than the actual modality just yet. So I wouldn't want to write one off. For example, the the protein marker that was published in nature medicine that I showed has an incredible you know um

1:15:48

Pearson R coefficient. Um and that's protein. It's not methylation or metabolomics. that probably just speaks to the quality of the data set that they used and the scientific rigor and the artificial intelligence that they used. So I think all these modalities there's a place for all of them. There's no clear winner just yet. And um I think feedback to myself is to have you know it would be cool to have a head-to-head comparison to to address your point. But thank thanks for the great question.

1:16:11

Okay, let's do quick quick questions here. Quick question. I was going to say uh there is a winning modality and that is the test of your functional fitness. how heavy you can lift, how fast you can run, how far you can run, right? So, I think a lot of from biohacking standpoint, I'm a biohacker. A lot of people have given up on all the these biological age tests and said a test of your functional fitness is actually what's going to really test tell you how long how your health span is for your life. Absolutely.

1:16:40

Fine. Great talk questions and quick answers. Yeah. Uh you were giving trials on the drugs. What drugs are you actually using and why are you using them? Oh, me personally? No, no, no. In the research, you said that you guys are doing clinical trials on the drugs and what age group are you actually testing in that? Oh, yeah. Uh, quick answer. Yeah. I I don't have all the specifics, but uh some of them are lifestyle interventions like a calorie restriction and it shows that calorie restriction, you know, is correlated positively with metabolic outcomes. There's one on rapamy um uh and uh you know similar I I don't know all the metrics off hand but uh and actually I couldn't find the the actual article um in time to post like a

1:17:22

nice figure from it. Um but there's kind of other uh agents that people are trying. Another one is um plasma. So giving plasma from a younger person like a blood product from a younger person into an older person as well. Very science fiction but you know we wonder if it actually helps people with um Alzheimer's disease uh too. And then others are just like good oldfashioned like diet interventions. So it turns out diet interventions, you know, just like your mother would have told you work work work super well and others are exercise interventions. So I don't have all the specific stuff for you today, but I think it's an excellent question and a great area of further study that needs biomarkers to understand.

1:17:57

Hi. So my background is in nutrition and medicine and one question that I had is how do these biomarkers actually drive concrete interventions something as simple as diabetes and weight loss that you mentioned behavioral change is hard and we know that the biomarkers don't necessarily translate into the outcomes. So my question is rather than predicting outcomes what do you think will actually change outcomes? Is there a biomarker that you foresee in longevity that we can actually concretely use to change something or can you just share something?

1:18:29

Yeah, I I think that's a fantastic question and you know um coming from a a dietitian point of view, you have a lot of expertise that as a doctor, you know, I I don't have and Oh, you're Oh, you're a doctor. Oh, you're Dr. Oh, sorry. Okay. So, uh but uh uh not a cancer doctor, probably a doctor that knows a ton about diet. And so, I think that's really wonderful. And um I think uh uh you know what I'll say is that um I was you know part of my my venture life is I was talking to a company recently uh that is actually looking at you know things like your your lean body mass and making biomarkers for that. Um I think really it comes down to a few things you know one is that the big picture from a pharma point of view or from like a science point of view is that these biomarkers will allow us to kind of instantaneously know outcomes in

1:19:11

clinical trials. But for patients, they allow patients to really know over time the impact of an intervention. And so instead of, you know, you walk into your doctor's office, it's some old dude and they're just like, you know, quit smoking, like, you know, eat better, stop getting Uber Eats, like the classic things that you always hear, you know, what if your like metabolic age or your protein age actually decreased in the 3 months where you quit smoking. Uh, and what kind of feedback and and and what kind of psychologically uh would that do to a patient if if they actually saw that? So, might there be some like feedback loop to a patient? But I think there's nothing more important than how a patient actually feels. And so, to our amazing photographers's point, like those actual physiological metrics, like how much can you lift? Can you run um V2

1:19:53

max? Or if you're an older adult, like can you actually spend time with your grandchildren? Can you lift a child? Right? Um can you take part in family outings? Can you travel across the country to make it for Thanksgiving? These are the things that I think biomarkers fall short of and struggle to quantify um but are so important. And so if a patient has feedback, they feel better and it's all coherent. I think that that would make a lot of sense. And thank you doctor for your question.

1:20:18

Anyone else? This is my question. I'm Jen. So I'm curious uh what's next in your research or do you share? Yeah. Uh yeah, a complex question. So I'm you know uh you know in my in my I guess my academic research I'm I'm researching with single cell RNA sequencing lung cancer and I'm trying to figure out which immune cell populations are mediating different outcomes within uh lung cancer treated by um amunotherapy.

1:20:54

So continuing on from my PhD research um but I think you know in the broader spectrum of things and my broader interests um are in healthcare and AI are in biomarkers um and uh in in all these things together. I'm actually particularly very interested in how large language models can read the clinical chart and predict patient outcomes and help us have kind of a large language model um uh you know health score uh that's that's read you know figured out by reading a chart instead of having a human doctor read it. I think that that will give us a lot of uh insight and a lot of markers about how long patients will live. There's been some early work on large language models predicting from a patient's record um in the cancer uh context like

1:21:34

how long somebody you know will live and and and I wonder if it might help us guide therapy selection um in cancer itself. So I think things with the large language models and the patient record are very provocative, very exciting. I'd like to kind of mesh that in a multiomic way with traditional things like the immune system, you know, blood draws, um, and figuring out the best chemo or amunotherapy to use in a cancer patient.

1:21:58

Those are some of the things that really excite me. I have one followup. And how would you want your how would you want your research to impact health care policies or health insurance here in the US or? Yeah. Yeah. Great, great, great question. again uh yeah I was born in Canada but uh what what I would say I went to school here and um done you know medical things here as well uh what I'll say is that um I want uh I want um insurers uh to be incentivized the same way that I want incentives to be aligned between patients insurers um and doctors I want us all to be in the same boat I want us to invest in things that make real measurable impact in health span

1:22:39

and lifespan and quality of life I I want my patients to have more good years on planet Earth. I want them to be able to go to Thanksgiving to spend time with their children. And I want insurers to pay for things that allow for those incredible moments um that give us joy in our lives. So, I want evidence-based effective things that we can measure with biomarkers to be invested in by insurance. Great tagline. All right, last question. Three, two, uh, okay, quickly, this is a quick question. Um, so for in general for precision medicine or u

1:23:20

personalized medicine, what do you think of uh DNA tools like websites which will let you uh input your raw data 23 and me and give you risk scores for various categories. uh one was incub but but there have been other I I think fantastic question extremely early days basically you know I was I was asking my attending you know one of the other days like why aren't we just genetically sequencing everybody part of it is you know a money thing it's very hard for patients to afford that to pay privately or for it to be covered by you know a government insurer by um you know by a private insurer um and so I think part of it is political part of it is

1:24:01

Um, and you know, part of it is that doctors, you know, we it's it's uh it's really new to us like correlating genetic sequencing outcomes with actual patient outcomes. But, uh, it's something I'm extremely excited about. I'm excited to learn more about it. And I think that we will come into a future where actually every patient is genetically sequenced um, when they enter into the clinical context. that first touch point when you enter into a family doctor's uh office or you enter into the uh the emergency room. I think that your blood will be sequenced uh uh eventually one day and it'll give us a tremendous amount of insight. So that's something I'm extremely excited about uh in the future but maybe not quite in the clinical context just yet. Thank you.

1:24:42

All right, thank thank you all for coming and let's give our people a round of applause and continue to enjoy. Thank you. [Applause]

1:25:07

Thank you. Thank you. [Music] Yes, I'm in New York, [Music] but I will. Yeah, for me, you know, I'd love to just

1:26:02

[Music]

1:26:12

kind of like situation [Music] environment anyway. Here's how we try and address it. Congress to bring people that can bring a very very busy

1:27:03

time. Thank you. We did enjoy. [Music]

1:27:26

[Music] Hey, you too. Stop. [Music] One, two. Sorry, I just moved this guy. One more. One more. One more. One more. Sorry. Sorry. [Music] I got [Music] [Music] [Music] my like it's just like kind of [Music]

1:28:51

workout today. Usually I do my workout. [Music] No.

1:29:23

[Music] I guess there was one down

1:29:47

between me. [Music] I don't really know.

1:30:20

What are you saying? [Music] [Music] [Music] [Music] I mean, they look pretty. I don't know. It's because they have a I think it's like a [Applause] Yeah, they're

1:31:19

very this is just something like just to hang on. So I found it off because I don't want to like come into [Music] last

1:31:57

minute. You know what? [Music] Yeah, I guess not. Not exactly. [Music]

1:32:39

You know guys,

1:33:08

[Music] [Music] You want to learn?

1:33:47

Congratulations. [Music]

1:34:00

[Music]

1:34:11

[Music]

1:34:26

[Music] [Music]

1:34:51

[Music] [Music] This is happening.

1:35:18

[Music] guarantee no life. [Music] [Music] But I didn't get [Music] Where's the

1:36:38

[Music] [Music] I don't know what I learned a lot about [Music]

1:37:18

[Music]

1:37:48

[Music]

1:38:01

He was like six. Oh my god.

1:38:14

[Music] take a lot of drug alcohol.

1:38:46

looking at me.

1:38:59

[Music] Oh, I just like a [Music] Okay. [Music]

1:39:24

[Music]

1:39:37

[Music]

1:39:57

I like to see [Music]

1:40:13

[Music] I think you guys like routines when you're kind of using I think. Yeah. [Music] Yeah. Hey guys, [Music] No, I did this. [Music] Are [Music]

1:41:26

you going You

1:41:44

might know. [Applause]