Peter Fedichev, co-founder and CEO of Gero, is a relative newcomer to the field of geroscience with a background in physics and not in biology. However, Peter has firmly established himself and Gero in the longevity landscape by twice publishing in Nature, entering a lucrative collaboration with Pfizer, and proposing a new aging-related theory. We discuss this theory in this interview, which was taken during the recent longevity biotech conference in Montenegro.
You are one of the few people who came to the longevity field without having a background in biology. How did this happen?
Well, the official, tidy version of my story starts with a childhood fascination with biology, but in reality, I was torn between studying physics and biology. At the time of that decision, physics seemed like the safer bet to me.
So, you went for the more foundational discipline, right?
Exactly, and I do love physics. I’ve been immersed in studying physics of complex and strongly interacting systems. If you think about it, isn’t life just like that: complex and strongly interacting? I believe that biology is becoming an exciting area where the principles of physics can be applied.
It wasn’t until later that I found inspiration in certain animals that exhibit what we call negligible senescence, or aging so slow it’s almost unnoticeable. This concept clicked with my physics background. I began to wonder if it was possible to create a physical model that could explain why some creatures age quickly, and others don’t seem to age at all.
I thought that maybe aging and “non-aging” could be two outcomes of the same biological system, just with tiny changes in how it’s regulated. If that was the case, we could potentially engineer ways to make aging animals non-aging. To me, this seemed like the most intriguing question in biology, perhaps even in all of science, and I decided to take a crack at solving it.
Just to clarify: although we know that some animals age much slower than others of the same size, whether their senescence is truly “negligible” is an open question, right?
Absolutely, and perhaps those creatures do age, just at an incredibly slow pace. Consider the naked mole rat. From what we’ve discovered, it doesn’t experience even one mortality rate doubling during its 40-year lifespan. In comparison, humans experience five such doublings during the same period. I’d be quite happy if we could reduce that to just one doubling; it would grant us an additional 200 years of life.
How was Gero born?
About a decade ago, we were getting serious about these questions. Data from animal studies was pouring in, and we felt like we could use this data to create predictive models, similar to those used for weather forecasts or financial market predictions.
Once you have these mathematical models in hand, you can perform thought experiments. You can ask: “what would happen if we changed this variable or that one?” So, we aimed to tackle the physics behind the aging rate and the mortality doublings to see how these factors could potentially be altered by a medicine.
We started to reach out to biologists like Robert Shmookler Reis. We conducted transcriptomics of incredibly long-lived worm strains. At Gero, we’ve gathered a range of data on worms that live anywhere from one to ten times their natural lifespan due to various interventions.
We’re talking C. elegans, right?
Exactly. We began with animals that have a highly flexible lifespan. With C. elegans, it’s possible to directly change the aging rate, offering an immediate link between theory and experiment. From there, we transitioned to working with mice. We collaborated with Andrei Gudkov and Brian Kennedy, began collecting human samples, and expanded our data generation and analysis techniques, including longitudinal proteomics.
Did you find biology more complex or less complex than you had thought?
It’s been a steep learning curve, to be honest. We initially thought it might be simpler, but we also believe that our skills complement those of biologists, who are often inundated with the volume of data they gather.
In the world of engineering sciences, we rely on statistical physics and mechanics to decipher complex systems. I often use this analogy: the steam engine was invented before there was definitive proof of atoms and molecules’ existence. This is statistical mechanics in action.
There are plenty of instances in physical sciences where it’s fundamentally challenging to understand how all parts of a complex system function together. Some argue that it’s simply a difficult task; others believe it’s impossible due to the sheer number of variables. But, thankfully, where one stands on this issue often doesn’t matter because detailed mechanistic knowledge isn’t always necessary for practical applications.
Like weather prediction, right?
Absolutely, or like hydrodynamics, which is used to forecast fluid flows and weather. To dig a bit deeper, there’s a long-standing debate in physics and the philosophy of science. Some, like Richard Feynman, are absolute reductionists, believing that if we fully understand all the fundamental forces, we could compute everything from chemistry to biology, psychology, society, and all other aspects of existence.
In contrast, Nobel laureates Philip Anderson and Ilya Prigogine, who studied complex systems in chemistry and condensed matter physics, noted a fascinating concept, which is highly relevant to biology. In his impactful paper “More is Different,” Anderson noted that as we move from one level of matter organization to another, new properties emerge. These properties don’t exist on a micro level but do on a macro level of a system.
Interestingly, these emergent properties often don’t depend much on microscopic details. This means that we can’t easily predict macroscopic properties of materials from microscopic details such as chemical structure. Consequently, we need to develop a whole new science for each complexity level in nature: chemistry is more than physics, biology is more than chemistry, and so on.
The concept of emergence is one of the most intriguing scientific developments of the last 100 years. In terms of biology, it may imply that medicine is more than just molecular biology. Extensive knowledge of individual genes and pathways may become irrelevant when new, large-scale properties like aging or chronic diseases emerge from interactions of biomolecules or cells.
The loss of information associated with moving up across complexity levels could explain why drug discovery is so challenging. We identify and administer drugs at the molecular level and anticipate results at the system or organism level. This might be the reason why it’s hard to find effective treatments for many diseases and why many modern drugs developed against a particular molecular target fail in human trials.
This line of reasoning is what I wanted to introduce to biology. I might not have had extensive knowledge of biology, but I knew from various textbook examples that the slower the process you’re trying to understand, the less relevant most of your mechanistic knowledge becomes. In other words, slower processes tend to be simpler and more universal, and there’s a solid set of procedures to describe the dynamics of these slow changes. In physics, when you can leverage this universality, it’s incredibly powerful.
Aging is the slowest process in an organism. If this approach is applicable at all, it should first work for aging and likely for chronic diseases. I thought that this perspective could give Gero a competitive edge.
Yes, Gero. It’s been around for a while, everyone has heard about Gero, and yet many people don’t know what you do. What is your model?
Essentially, our objective is to first understand and then halt human aging by basically engineering negligible senescence. We initially got support from people who appreciated this goal, believing that if we succeed, it could become a profitable thing. With time, we’ve gained more interest from traditional pharmaceutical companies. Take our recent partnership with Pfizer, for example. Since almost all diseases are age-related, our advanced understanding of human aging could help develop new drugs to combat ‘conventional’ diseases.
So, it’s a moonshot, but you have investors who share your vision?
Absolutely. We’re fortunate to have investors who’ve stayed with us, as well as newcomers. These people continue to support us, enabling us to manage Gero in an increasingly business-like fashion, but our primary goal was always to understand aging and develop powerful anti-aging interventions. Most people in this field accept that a successful intervention must alter the rate of aging.
How do we even measure the rate of aging in humans?
Great question. The most effective definition of aging I know is that it’s the exponential increase of mortality. In humans, the risk of death from all causes doubles every eight years, but mortality is a population-level trait, and we need to connect these population figures with measurements in individual organisms.
This challenge is central to the aging biomarkers field. They want to measure something in you today that correlates with the characteristics of population mortality trends. This is why we began examining longitudinal datasets that contain multiple measurements from the same person. We wanted to understand how physiological parameters change throughout a person’s life.
We couldn’t do this with nematodes, since measurements would likely kill them. So, we turned to data from mice and humans. We started with a large public Mouse Phenome Dataset and added retrospective data from Andrei Gudkov’s years of mouse aging studies. We also procured a vast human dataset from a diagnostic company in Moscow with data points from individuals who took multiple blood tests over the company’s 20-year history.
When we started comparing mice and humans, we found something intriguing: although both species show an exponential increase in mortality, the dynamics of individual markers in humans and mice are completely different. Humans are not just bigger mice.
This discovery made us revisit our theories and rethink everything. We had to face the fact that humans are very different longitudinally. In mice, we see mortality increase exponentially, but so do biomarkers of aging. This pattern of exponential codependencies is everywhere.
Markers of inflammation, such as c-reactive protein, IL-6, and others are rising exponentially in mice. So is the burden of senescent cells. The exponential rate matches the mortality acceleration. This means that mouse aging is simple: we observe an exponentially accelerating breakdown of the organism’s state.
In humans, however, we know that after the age of 40, our mortality doubles every eight years. So, we see five doublings of mortality, a total 30-fold increase between 40 and 80.
Even before we hit 40, signs of aging begin to show. Are these just more subtle aspects of the same age-related curve, too gradual for us to measure accurately?
It’s clear that not all aspects of aging in humans follow an exponential pattern. Our facial features do shift with age, but not exponentially. For instance, the space between our eyes might increase, but it doesn’t multiply by fivefold by the time we’re 80. Just picture that!
If you were to chart various human characteristics over time, you would likely find two distinct patterns. Many aspects change in a straight line, getting more varied with time – showing that their change is random. Then there are those markers that change faster than a straight line – hyperbolically. If you were to extend these lines, some would reach an infinite point at around 120 years – the current maximum lifespan. Interestingly, these are the same markers that show an exponential increase in mice.
This subtle but qualitative difference (hyperbolic vs. exponential) already shows you, even without any interpretation, that aging in humans is very different from aging in mice. For instance, mice experience an exponential rise in death rates until their average lifespan, then it plateaus. But in humans, once death rates begin to rise exponentially from around 40, they keep doing so beyond the average lifespan. In simpler terms, mice and humans age quite differently, and we need a theory to explain this.
Scientists are beginning to realize that mice may not be the best models for studying human aging.
Yes, our work aims to highlight exactly how human aging differs and identify aspects of it that can be studied in mice or even in C. elegans.
Basically, we find that severe late-life health issues in humans resemble aging in mice quite a lot. In humans, aging and frailty are separate – an older person might not necessarily be frail and vice versa. Doctors and aging researchers use ‘frailty indices’ to measure aging in patients – they tell us about the burden of disease, not about demographic trends. In mice, it seems, aging and frailty go hand in hand – perhaps because frailty sets in so rapidly that it looks like the aging process itself.
On the other hand, in humans, frailty typically sets in only in the last decade of life. Before that, other aging-related changes are happening, changes that we don’t see or don’t have time to observe in mice.
Let’s go back to your theory and explain it in the plainest terms possible. So, you are saying that there are two types of aging: mice-like aging and human-like aging.
We propose a simpler classification. Let’s do a thought experiment. Imagine a living system in a balanced state, and you introduce an error into this system – for instance, by removing a few molecules. This action damages the system, at least temporarily. Now, two scenarios could play out.
In the first, the absence of these molecules triggers a chain reaction of errors that escape the system’s repair mechanisms, causing more and more errors. This cascading effect could cause deviations from the balanced state, increasing exponentially until the system disintegrates. If we were to draw a comparison with physics or engineering, these systems would be dynamically unstable. When mortality increases exponentially, it’s natural to assume there’s an exponential instability in the system.
In the second scenario, the system’s error repair mechanisms efficiently correct errors before they can cause more errors. It’s like having a system where the basic reproduction number of errors is less than one. In such a case, errors occur but don’t multiply, allowing the system to survive for a long time without showing signs of aging.
Interestingly, we see that similar organisms, with slightly different parameters of regulatory interactions, can have very different aging outcomes. A minor change to the parameters that govern error repair systems can determine whether or not errors amplify exponentially. This was our initial concept: similar species, like mammals, share much of the same genetic makeup. We all have genes that, in theory, can repair almost any kind of damage, but different species have evolved different approaches to fixing these errors.
From our observations, there are usually good reasons not to fix every error. In evolutionary terms, energy can be spent on growth or on repair. Too much energy spent on repair slows growth. This is why there are no worms living ten times longer in the wild – as explained to me by Robert Shmookler Ries. Long-lived mutants develop and reproduce slowly, which washes their genes out of the gene pool.
This is the story of the first generation of intervention in worms. Later, we did find interventions that significantly increase lifespan in C. elegans (but not tenfold, I think) without slowing development and reproduction too much. But, in general, this trade-off between development, reproduction, and repair is real.
Yes, in most species, the rate of development and aging are interconnected. If too much energy is invested in repair, development slows down. Social animals often develop more slowly.
And evolution is making this choice for us.
Yes, and evolution doesn’t really have too many options to play with. It has to balance between growth and repair, especially when considering energy conservation (this idea is due to Geoffrey West). If there are too many individuals competing for resources, a viable strategy might be to try to outgrow and out-consume them, but this might also mean investing less in repair, which can lead to less stability over time. But if this instability doesn’t bring you down before you reproduce, evolution doesn’t really mind.
In essence, competition drives us all to the edge of instability. This is something that was first noted by Stuart Kauffman, a medical doctor turned influential physicist, who observed that all living creatures produced characteristics of regulatory systems operating on the edge of stability and instability. His work was a significant inspiration for us.
Basically, competition makes you live on the edge.
In fact, it can push us over the edge. To compete effectively, it might be a good idea to disregard repair and allow for exponential disintegration. This is the “anti-engineering” solution that nature often opts for. Nature doesn’t mind if you’re unstable. If you choose to grow quickly and then begin to disintegrate, but manage to reproduce during that functional period, that’s all evolution needs from you.
But if you don’t face this kind of pressure — if you’re at the top of the food chain, survive to a certain age, and continue to reproduce for a long time, which is crucial — then the gene pool could become dominated by genes that promote longevity, simply because a longer life allows for a more extended reproductive period.
I’m not a biologist, and that’s why I generally avoid these discussions in my papers, but I believe that humans are caught in this unique situation because of menopause. On one hand, we’re at the top of the food chain, but on the other, we stop reproducing way before the end of our lifespan, which is not too common in nature.
Both humans and naked mole rats are social animals, but we’ve developed two very different social structures. In naked mole rats, it’s like insects, where one “queen” reproduces throughout her entire life. So, if genes promoting lifespan extension enter the gene pool, they have a high chance of sticking around and accumulating.
In humans, older women often help their daughters to produce offspring. This could be why evolution might favor switching off reproduction in older women to establish a balance. There is strong evidence that just a few hundred years ago, when human living conditions were harsh, having your grandparents alive significantly increased your chances of reaching reproductive age.
This is often referred to as “the grandmother hypothesis”.
Yes. While this may have been instrumental in shaping our social organization, it has disrupted the connection between longevity and dominance in the food chain. We’re a dominant species that could have allowed for more longevity genes to enrich our gene pool. I believe this is still happening, but a bit slower and for different reasons.
Humans are living longer as we’re social creatures, and our society is ever-evolving, growing increasingly complex. This is why it takes more time now to mature and adapt socially. So, nature might be subtly extending our development time by enhancing our repair mechanisms.
People now appear younger at 50 than they did a hundred years ago, even after accounting for advancements in medical technology and other factors. But due to menopause, we can’t have a situation where longevity genes aggressively accumulate in our genome.
Most of this is irrelevant to mice. We both agree that mice aren’t the best model for studying aging, but what other options do we have? Fortunately, we’re living in a golden era: we have more medical and molecular data on humans than on any other animal. So, we might not even need to study animal models as extensively as we once did.
Yes, we are limited in experiments in humans, but if we can achieve a similar level of understanding of human aging and disease dynamics as we have for weather or markets, we could learn a lot even without direct experimentation.
This is where I see great potential at the intersection of physics, biology, and machine learning. If you understand the dynamics of your subject, you don’t need to conduct countless experiments because you already have a good idea of the outcomes based on your understanding of data generation. Once you know the dynamics, you can run simulations. This is how airplanes and weapons are designed, and hopefully, it’s how drugs will be designed in the future.
That’s why we’ve started building models based on longitudinal data. We’ve already developed such a model for mice, which we published last year. Two and a half years ago, we published another model using longitudinal human data.
Today, thanks to our collaboration with Pfizer, we have models that have been trained on tens of millions of medical records. These records distinguish between aging and chronic diseases, allowing us to see how aging impacts chronic diseases. With these comprehensive models, we have access to superior aging phenotypes from genetic studies to inform our aging research.
Let’s step back and examine the foundation of your theory. I’m going to try and simplify it, and you can tell me if I’m on track. We’re discussing two distinct types of aging. Humans are superior at self-repair compared to mice. Our robust repair systems only start failing much later, when aging accelerates.
But this means we can’t effectively use anti-aging treatments before this acceleration kicks in, as our repair systems would counteract their impact, constantly restoring the body to its slow-decay equilibrium, right?
That’s right. There are animals that maintain stability and others that rapidly deteriorate, like mice. When humans reach full development, we arrive at a stable state. Both negative influences, like smoking, and positive ones, like taking anti-aging drugs, don’t significantly alter this balance. For instance, consider smoking – even with its harmful effects, it only reduces lifespan by around 5 years. This impact, less than a 10% effect on lifespan, is comparable to the effect of gender.
This is what stability means. It’s hard to modify human lifespan. But this stability isn’t infinite. Within a certain range of changes, there’s a force that restores balance. But if you push too far, you’ll end up in an unstable state and begin to deteriorate.
What we noticed is that although we mature into a stable state, this stability gradually weakens as we age. Just as the size of our nose and ears change steadily with age, so does our recovery force. We mapped this decline as a function of age in longitudinal data.
Two consistent observations stood out. First, the variation in fluctuations increases, suggesting that the ability to maintain balance weakens. Second, the force that restores balance after a stress weakens. It appeared that the only thing changing was the degree of balance, with the restoring force growing weaker, leading to larger fluctuations. This suggests that in addition to the typical markers of aging, something else is gradually reshaping everything within us, from the shape of our nose to our inherent balancing force.
This dominant aging process in humans is linear, which is interesting. Compared to mice, nature has already done most of the work stabilizing our internal systems. We’re left with this slow, linear process leading to eventual instability late in life, while mice experience an exponential process almost immediately after birth, shortening their lifespan significantly.
You have said that aging-wise, we are like naked mole rats for most of our lives, and that’s probably true. We may be frustrated about our limited lifespan, but it’s actually already very impressive for an animal of our size.
Exactly, we and naked mole rats share many similarities. Perhaps they too experience this linear reshaping of their regulatory systems, just at a slower rate than humans. Some people take a negative view of this, but my theory’s positive statement is that nature has already done most of the bioengineering required for life extension. We are stable, so we don’t experience exponential aging for most of our life. We only have this linear decline to contend with, and if we could slow it by half, our lifespan would double.
Another part of your theory suggests that this slow linear aging is like an arrow of time, going in one direction and very hard to reverse. So, we might be able to achieve negligible senescence (to slow the clock), but it’s much harder to achieve true rejuvenation (to turn the clock back).
Yes, it’s a practical theory that provides a systematic approach to identifying factors that control the rate of aging, removing the reliance on fortuitous experimentation. If our theory is correct, certain factors would behave in predictable ways, and we should see this reflected in experiments.
Unfortunately, if we’re correct, this slow, linear aging is not a single process but the cumulative effect of many unrelated processes. There are countless variables that could go wrong. A molecule could change shape, methylation states could alter and fail to revert, mutations could occur – there are simply too many possibilities.
Each minor change might seem insignificant, but if you wait long enough, over a near 100-year lifespan, the compound effects of all the accumulated minor insults changes every variable in your body. In response to this stress, the recovery force gradually weakens. Interestingly, for some processes, the recovery force strengthens. Certain diseases don’t affect the elderly. But naturally, the weakest link is what matters. The system that breaks down first, becoming unstable, is what determines our lifespan.
If we’re right, and this slow aging is the result of many uncorrelated processes rather than a single one, it means that it’s extremely challenging to develop one or even a few drugs that would have a significant impact. Different cells experience different problems, so any treatment would need to be highly specific.
But it’s not exactly impossible, even with existing technologies. We’ve learned to target our therapies pretty well, even on cellular level.
Theoretically, it’s possible. It would require less energy than what you get from a daily cookie to clean up all your DNA. It’s not about energy, but rather about information. Take, for instance, trying to swat an irritating mosquito. You exert a lot of energy, but that’s not the issue. The problem is that you often miss and hit yourself instead – you lack the necessary knowledge about the mosquito’s position and can’t quickly act on that information.
You’d need to go into every cell, determine what’s gone wrong, and then apply a highly precise intervention that targets only what you need and leaves everything else intact. In physics, theoretical things like nano-devices acting on individual elements in complex systems are referred to as “demons”, like the famous Maxwell’s demon. These are like the nano-robots we were promised a long time ago. If we could do that, it’s not impossible. Nature’s laws don’t forbid it. But it’s a monumental task.
If I had two projects based on current technology, one to slow aging and another to significantly reverse it, my money would be on the first one. Why? First, because my background in physics tells me so. Second, we see creatures like the naked mole rat and other animals that have apparently slowed the pace of aging by an impressive degree.
So, this is a hypothesis, albeit an interesting one. Do you believe it requires more supporting data?
Actually, what we need now isn’t more data but more experiments. The theory has been built from data. Theories allow for predictions. For example, there’s a recent paper by João Pedro de Magalhães that, like our work, showed that the variance of the “epigenetic noise” increases with age. He proposed examining this additional biomarker of age – the variance.
I’m predicting that some drugs could reduce the variance of the biological age at any age and, in turn, extend lifespan by making biological networks more robust. That’s a prediction, and that’s the power of theory. Even when working on the wrong animal but examining the right things, a theory can lead you a long way. You can observe things that are relevant to your theory and then attempt to apply those findings to humans.
Could things like organ replacement or epigenetic rejuvenation (cellular reprogramming) work around your theory?
Possibly, but to make a significant impact, you would need to remove a substantial amount of damage. It can’t just be a handful of cells. In my opinion, cell rejuvenation holds the most promise, at least theoretically, because it’s currently the closest thing we have to those theoretical “demons” from classic thermodynamics.
However, as we’ve seen from experiments, if you increase the dose in mice, you see a beneficial effect for a while, then it peaks, and then it decreases. The peak effect isn’t more significant than interventions targeting other hallmarks of aging.
How was your theory received by the scientific community? I know some people were skeptical and even displeased. After all, you’re sort of a buzz killer.
Overall, I think the response from the community has been positive. We’ve received help all along the way. Our theory lines up with what many in the field have experienced when testing longevity drugs in humans. More often than not, we see what we predicted: small, sometimes temporary effects that can’t outperform the effects of a good diet.
Yes, at the conference we’re currently at, people like Brian Kennedy have mentioned that some of their results are at least compatible with your theory.
I don’t want to come off as pessimistic because, like many others, I joined the field of longevity research hoping to achieve dramatic rejuvenation in humans. There’s also significant funding flowing into this field, and we must manage expectations accordingly. If, after all this investment, we end up with interventions that mimic the effects of calorie restriction, there will be a lot of disappointment. Basically, we risk undermining the field if, in five to ten years, we find that everything we have is no better than diet and exercise.
But clearly, you still want the field to receive ample funding.
Absolutely. I believe that having a flawed theory is better than having no theory at all because it saves time and money. When we promote the concept of longevity to outside investors, we’re building expectations, and we must be responsible in managing them. For instance, we should consider the effects of calorie restriction as a fundamental benchmark.
What is then your message to people who want to invest in longevity biotech?
The answer really depends on their risk tolerance. Many investors want to help people towards the end of their lives cope with specific diseases that currently have no effective cure. For instance, I commend what BioAge is doing. They’re repurposing existing drugs and rapidly progressing them through clinical trials to help people, especially the elderly, in specific situations such as post-COVID. In this scenario, there’s a distinct medical problem, a substantial number of people who need help, a potential for profit, but also a significant opportunity to gain experience with anti-aging drugs.
Likely, they will be able to enhance patients’ quality of life and provide returns for investors. This is an ideal fit for those investors looking for proven methods that align with modern medical philosophy. Some people will profit, while others might, say, win a Nobel Prize for something like senolytics.
With these interventions, we can hope to extend the last period of life by maybe 10 years for people who are already in bad shape. This is what compressing morbidity is. We shouldn’t downplay its importance, it’s substantial, but outsiders often have different expectations. Our books and speeches lead them to believe that we can halt or even reverse the aging process in someone in their 30s or 40s. But this won’t happen with the current generation of interventions.
So maybe, Gero’s proposition is: “If you carry on this path, you’ll have to wait at least 10 years before realizing these approaches don’t work as you’d hope. Then you’ll have to wait another 10 years for new strategies to be developed. By collaborating with us, you might save 10 years.”
Your theory suggests we have a better shot at maintaining an 80-year-old in reasonably good health than rejuvenating a 50-year-old. Would you personally take this deal to stay 80 for a long time?
If I’m still vigorous and relatively healthy at 80, then absolutely. My mother is nearly 80, she’s not in perfect health, but she wants to stay around, maintain her independence, and help her grandchildren. At the very least, we should have this option, and then it becomes a matter of personal choice.
