The trajectory of our health does not follow a straight line, says Michael Snyder, Ph.D, the Stanford W. Ascherman Professor in Genetics at Stanford Medicine. And his work of the past decades has delivered the data to prove this, using wearables, genomic testing, questionnaires, MRIs, and more to collect health metrics that illustrate just how uniquely our bodies age.
During Snyder’s session at DOC 2025, “Transforming Healthcare Using Deep Data and Remote Monitoring,” he walked the DOC community through his research, which is based on hundreds of data points. His lab examines details from heart and hormonal health to inflammation and oxidative stress, then calculates a biological age, as opposed to the years traveled around the sun. He considers this process more accurate than methylation clocks (which examine biochemical changes over time) as the results he uncovers are “actionable,” he said.
“You use AI to make very specific recommendations, mostly around food you should eat,” he said.
This individuality of aging is “a big theme of our work,” illustrating his point with data from his lab. One experiment revealed very personalized responses to taking the popular supplement drink Ensure, where some participants saw inflammatory spikes, and others saw reductions. Another tracked reactions to glucose, from sources including rice to pasta, and then uncovered people’s unique subphenotype that came from this data.
Snyder hopes that one day everyone’s aging types and genome sequences will be tracked, even from before birth, to uncover clinically actionable data to better manage people’s health.
You can hear more of Dr. Snyder’s insights in our video from DOC 2025 or read our lightly edited transcript below.
TRANSCRIPT:
Dr. Michael Snyder
It is a pleasure to be here. Our lab does a lot of work with, deep data and remote monitoring. That’s what you talk about. But, mostly emphasize the of our work with nutrition, which has not been covered a whole lot yet at this meeting. So many of you may know a number of years ago. We started this project to start collecting very, very deep data on people where we would collect their blood, urine, a stool, and other samples and actually do these deep omics profiles meaning sequence or genome, do transcriptome, proteome, metabolome problems, along with wearables, clinical tasks, questionnaires, all other kinds of data.
We did this on people while they were healthy with the goal of trying to understand what a healthy profile was, not a single profile. How does it change over time? Can we use advanced technologies like genome sequencing, wearables? And now MRI to be able to better manage people’s health, and actually that does work.
It started out a small group, people 109, and just from the first three and a half years of running this study, we had 49 major health discoveries, and nearly half the people learned something pretty important about their health. Some of these were a big deal and covered all these areas hematology, cardiovascular, metabolic. I should point out that these are all found pre symptomatically that folks do not have symptoms. They’re revealed by these deep data profiles. Some of them were a big deal like lymphoma. We call it early pharma to pre-cancer is to people is here is heart issues. Some were found by genome sequencing. Some were found by wearables. That was another class of no one technology found these. Sometimes it was the omics. Sometimes imaging what have you.
We’d like to think we’re getting a much more complete picture what’s going on with a person’s health. That’s what reveals the these things. Then the other thing we discovered is because we’re tracking people over time, we can see how they change with time, which we interpret as aging. Remember, we’re taking healthy time points from folks. What we discovered is that everyone changes differently. So person number two of their their top biochemical pathway that changes is our cardiac hypertrophic signaling pathway, their cardio age. Or if you well we later learned their state two hypertensive. The person on the left has our immune system metabolic system and other parts changing. They’re a pretty typical ager.
We classified people into what we call aging types, aging patterns. At the time, nerve for kidney, liver, metabolic and immune and the people all the way on the right, they have let’s see why is this point. Are they actually are aging all four. This person’s not a kidney aging but aging. Now the three this person’s a kidney ager not much. So again, and there’s clinically actionable data associated with these aged types of people. In fact, some of them did actually intervene and they improved their markers. I’m not saying they got younger, but they did, in fact, improve their markers. The other thing we decided to do is because we’re trying to people over time, we can say, well, how are people really aging or certain times more, or the times when more things are happening than other times? Are we just aging linearly if you will? The answer is very few biochemical processes are aging linearly. Most are actually showing. So it changes. Rhere are two periods where there are big bursts of changes. One in the 60s which we expected people’s immune decline. Other things happen as people hit their 60s, muscle mass loss and wrinkles, for example.
Although we found many more things, the 1 in 44 was not necessarily expected, although anecdotally, a lot of people did actually go through have had these things. And there we discovered once again, there were changes in muscle, but lipid metabolism shifted along with alcohol and caffeine metabolism, which might explain certain shifts people experience as they hit their 40s. The other thing we’ve been pushing very hard on is now is this nutrition. How does, food really impact your health? I am a believer is probably many people here you are what you’re eat. We’ve been running a series of studies I’ll tell you about few of these. Here is one on fiber where we all know fiber is good for you, but fiber is very heterogeneous.
It’s, as long chain, short chain, positive, negative, hydrophobic, hydrophilic. It’s like calling all animals the same, cockroaches and humans. We’re all the same, right? I don’t think so, from a human perspective. So the point is, it’s useful to know what these different fibers are. Which ones have what phenotypes or effects. So we tested two inulin and arabinoxylan as supplements where he gave increasing doses to folks. It’s a smallish group of people. We gave them these two popular fibers. Arabinoxylan is in husk and Metamucil, for example. And inulin is in other things. We gave them either ten grams a day for the first week than 20 than 30, then a washout. Then they moved to the next fiber, and then later they were randomized. Then later they did the mixed fiber, which is supposed to be best. But the point is, at the end, there are hundreds of changes that occurred actually with each of these fibers. But the literature said depends. It was controversial. They might have positive effects on cholesterol or glucose, what have you. Neither. Overall, improved glucose, cholesterol was actually highly improved by arabinoxylan.
The system has won quite a bit, actually went down about 25% for the group as a whole at the highest doses. So that one did have a very, very positive effect. But even though it worked well on the group as a whole, quite interestingly, you will see differences at the individual level. And that’s a big theme of our work. Everybody is different. So for example, here’s the person on the left, the one in blue. Their, their textbook, their cholesterol goes down with the island. The one in orange. That person actually is not affected by a ravenously island, but actually their cholesterol goes down with inulin, the other fiber. But not the, not the blue person, if you will.
The point is, we do have these personalized responses. And so it’s it’s critical that you know, how this affects you, not just the group as a whole. We also do a lot in the area of remote monitoring. We do a lot with smartwatches, probably the most, you know, I where a lot of these things, like wearing my hearing aids, the sensors as well that my expose on how many you’re wearing a, let’s start with the smartwatch. Yeah. Nearly everybody. How many of you wearing CGM or have worn a CGM now? What a great crowd. All right, so I our claim the fame as we started putting these on so-called normal and pre-diabetic and discovered that a lot of them are actually spiking their glucose, some just as bad as diabetics. And we classified people into glucose types.
Then we went on Aaron Siegel, we went on to show that different people spike the different foods. It’s very personalized. And so this is the latest study we did where we had 55 people eat seven different carbohydrates that were in different forms. They did it in duplicate sometimes triplicate. Most people are like the one at the bottom on the left there, rice bikers. In fact, for many people, probably mouse rice is worse than ice cream, actually, in terms of spiking your glucose. So most people are like that. Interestingly, Asians spiked more than nine Asians, something I don’t fully understand. But there are plenty of other patterns to the one on the top left is a bread spiker. Next one over a grape spike, or in green, next one over a different shade of green, pasta. Next one over, potatoes were all different. Okay, what’s going on? We think this depends on your sub phenotype of diabetes. We classify diabetes type one or type 210% of type one, 90% of type two. But we actually think type two as many, many, many subtypes. Here’s why. Many different organ systems are involved. Liver, pancreas even your brain is big glucose consumer muscle many biochemical pathways. You heard about the GLPs, but there’s also insulin.
We started classifying people by their diabetic sub phenotype as well or glucose, this regulation in this case. So on the left is hemoglobin A1, C blue or a normal pre-diabetic, sorry yellow. And a diabetic snuck into our study. But we have many subtype from muscle one to resistant and beta cell defects anchored tonight GLP you heard about hepatic. And so where is this that you’ll see it’s all over the map. So if you look at this person here, the top one basically they’re muscle into resistant intermediate for beta cell normal for anchor ten and so on. But the one at the bottom has a severe incretin effect and other partial for other things. So we’re all different. You’re going to have combinations of these things as well.
We can actually now subtype someone just if they drink a shot of glucose while wearing a CGM, at least for two of them for the muscle and to resist some beta cell for it. We can tell because the patterns from your CGM, we’re all different in our patterns, and it has to do in part to the sub phenotype we have of of glucose this regulation.
So what what do you care about your sub phenotype? Well, depending on your sub phenotype determines the food your spike to. So if you’re Muscle IR resistant you’ll be a potato spiker and a pasta spiker. That’s the yellow and the gray one that’s third down, but not if you’re insulin sensitive. It turns out if your beta cell defect, you’ll also be a potato spiker. So knowing your sub phenotype determines what foods you should eat to better maximize your glucose. We can take this one step further. We can actually see people wearing a smartwatch while wearing a CGM, and there’s subphenotype than they do food logging, etc. and then we can build these AI programs to see not only what things you do, but when you do it and how that affects your your glucose dysregulation and other parameters.
I’ll walk you through a few obvious things. Eat your first big meal in the morning, your glucose is lower. The hemoglobin A1C starchy vegetables bad high at night later in the day, meal, if your big meals, then that’s trouble. Higher glucose, fruity carbs better. Most people don’t sleep enough. If you sleep more, you get your glucose down so you can start matching again what you do and when you do it. But you can also match it to your subphenotype. So if your Muscle IR resistant you should exercise more, and that’ll help improve that. I personally am diabetic. I have a beta cell defect. I lifted weights, gained 10 pounds of muscle mass by whole body MRI. Totally failed for my glucose control because I’m a beta cell defect, I don’t have a muscle insulin resistance. It’s not going to help me to build muscle mass. So knowing your sub phenotype determines what you should do.
I’ll just wrap up with one more thing. If I can get out of the slide here, hopefully. There we go. We’re doing a lot with micro sampling these days. These little drops of blood, you mail it in to the lab, and we’ll measure 7000 molecules these days. I know what it sounds like, but ours does work. And the point out of all this is that we can then start running fun experiments. Here. We had 32 people drink the shake, and they all responded differently. Here’s three carbohydrates. This person here is 13. Their carbs drop when they drink the shake like people over. They all go up. This one goes down, down. And formatically we can classify people into these Insurer types. I guess they are. And basically you’ll find that, brown is inflammatory markers. So this group of people, their inflammation goes down when they drink the shake for the next group over, it goes up. So same shake, pro-inflammatory on some, anti-inflammatory and others. We don’t fully know why that is. We have our theories. But you can measure it and actually do something about it.
We’ve commercialized this. But the point out of this is that by a simple test of 650 metabolites, you can actually see what patterns are going on on people, for 20 different, if you will, wellness categories, oxidative stress, inflammation, hormonal health, heart health, etc., even longevity.
The last story I’ll tell you then is that, it’s actionable information. You basically get aging type information so that like this person here, their biological age is older than their chronological age. We think it’s better than the methylation clock because again, it’s more actionable. And then from your aging type you can see well this person’s fine for inflammation. Not bad for metabolic but actually their heart age is two older. Then of course use AI to make very specific recommendations about mostly around food you should eat to better improve this. So this is what you can do to deal with this. This is Mike Snyder’s world. I envision a world where people get their genome sequenced, ideally before birth, and then together with molecular and wearable measurements, you can better manage people’s health. Everyone will have a personalized AI tracking system. Some of us already do. That’s it. Thanks.