Entropy in Aging: Notes From Two Days in Cologne

For three days in Cologne, physicists, biologists and bioinformaticians sat in the same room using one word for very different things. The word was entropy.

From 31 August to 2 September 2026, the International House of the University of Cologne hosted the Entropy in Aging workshop, organized by David H. Meyer, Maximilian Unfried and Sang Wouk Cho. Speakers included Andreas Beyer, Aubrey de Grey, Uri Alon and Ian White. Laura, Kiarash and I attended the talks on the first two days and skipped the workshop sessions on day three, which may even have lowered the entropy in the room by a measurable amount.

That variety was our most important observation. Someone coming from physics means something different by entropy than someone who analyzes data or studies cells. It shows how young the field still is: the role of entropy in aging is an open question, and even the terms are not settled yet. That made the workshop timely. The organizers describe its goal on the website:

Hosted by the Tandem Project at the University of Cologne, this focused gathering aims to advance entropy as a conceptual and quantitative framework for studying aging.

Program or chance

Behind the word lies an old core question of aging research: is aging a program that runs in the body, or the result of chance piling up over the years? Programmed theories see aging as a regulated process, much like development from embryo to adult. Stochastic theories see it as errors and deviations that nobody planned. In this debate, entropy is often associated with the stochastic side of aging. The idea is that random changes accumulate over the years and increasingly alter biological structures and processes. Whether and how these changes can be described in terms of entropy is a separate question.

Aging clocks gave the debate new momentum. They estimate biological age from measurements, usually chemical marks on DNA. Because they work so reliably, some see them as a hint of a program: if age can be read so precisely, there must be a regulated process behind it.

This is where a 2024 Nature Aging paper by co-organizer David Meyer and Björn Schumacher comes in. They simulated data in which the marks on DNA or the activity of genes changed purely at random. That accumulated chance alone was enough to build aging clocks. Applied to real data, these clocks estimated age and picked up differences, for example with smoking or eating fewer calories. So an aging clock does not need a program to tick. Accumulated chance, exactly what entropy is often associated with in this debate, might already explain why it runs so reliably. The authors explicitly do not rule out a program.

The model does have one requirement: a starting state from which chance accumulates. A 2021 finding by Csaba Kerepesi and colleagues fits this. In mice and humans, aging clocks show that the embryo gets younger during an early phase before its biological age rises again. The authors call this lowest point "ground zero" and see it as the moment aging begins. That provides a possible starting point the model needs.

One word, several meanings

If accumulated chance might explain why aging clocks tick, the question becomes: what is entropy in the first place? The question is not new. Back in 2007, in an article in PLoS Genetics, Leonard Hayflick argued that entropy explains aging. He also criticized how loosely many terms in the field were used. Nineteen years later, that problem is still not solved.

In everyday language, entropy often just means disorder. Specialists, however, mean different things depending on their field. In physics, it is about counting: how many ways can the particles of a system be arranged without changing the overall picture? The more ways, the higher the entropy. In information theory, it is about surprise: how hard is it to predict which outcome you will get? A coin that always lands heads has zero entropy. A fair coin has the highest entropy possible with two sides.

Aging research usually means the second sense: how unpredictable are the measurements? Then you have to decide what exactly you are looking at, for example marks on DNA, the activity of genes or the mix of cell types in a tissue. Depending on that choice, the same calculation measures something different. Two very different patterns can even give the same value. So an entropy value only means something once it is clear what it refers to.

Physics does not offer a ready answer either. You often hear that aging simply follows from the second law of thermodynamics, which says entropy never decreases. But that only holds for an isolated system. The body is not one, because we eat, breathe and give off heat. That lets us keep order inside us while the entropy of body and surroundings together still rises. A fridge does the same: cold inside, warm at the back, the electricity bill pays the difference. So the law alone does not explain aging.

Many perspectives

The speakers' perspectives were as varied as the meanings of the word. Four examples from published research show the range: from genes to the risk of dying to the question of how to measure whether an intervention works. Not all of these studies measured entropy in the strict sense. But they examine processes often associated with the term, such as random changes or a loss of coordination.

How genes work together

In a 2024 study, Ana Carolina Leote, Francisco Lopes and Andreas Beyer examined how genes work together as we age. Genes do not act alone; they regulate each other. Across eight human tissues, most of these links held up with age. The ones that weakened were mainly between genes that do different jobs in the cell, while others even grew stronger. So aging here does not mean that everything slides evenly into disorder. Above all, the coordination between different jobs in the cell declines.

The risk of dying

Omer Karin, Uri Alon and colleagues use chance to explain why the risk of dying rises so steeply with age. In their 2019 study, they found that senescent cells are cleared more slowly in old age than in young animals. These are cells that no longer divide but do not die either. Because they are removed more slowly, random fluctuations in their numbers persist longer and differences between individuals grow. In the authors' model, death occurs once these fluctuations cross a threshold. That reproduces the Gompertz law: from adulthood on, the risk of dying doubles at regular intervals, in humans roughly every eight years. In this model, slower removal and the fluctuations that follow are enough to explain the steep rise.

Omega-3 and aging clocks

Karsten H. Weylandt, who has researched omega-3 fatty acids for years, also spoke in Cologne. The DO-HEALTH analysis shows how carefully measurements in this area need to be read. In a post hoc analysis of 777 adults aged 70 or older, those who took one gram of omega-3 daily for three years showed slightly less aging on three epigenetic clocks (PhenoAge, GrimAge2 and DunedinPACE), roughly three to four months overall. That is a change in measurements. It does not show that these people will live longer in good health. After the Meyer and Schumacher study, it also remains open whether the clocks capture a program or accumulated chance.

Combined interventions

Joachim Krug, a theoretical physicist at the University of Cologne, looks at lifespan from a physicist's angle. In a 2018 paper in Biology, he and colleagues described a method to test whether two interventions work better together than expected or simply independently. It predicts what the survival curve looks like if both act independently, meaning how many animals are still alive at each point in time. In worm data, the interventions mostly acted almost independently, even where the usual comparison of average lifespan suggested a stronger interplay. Looking at the whole curve instead of a single average leads to a different conclusion.

What we brought home

In Cologne, fields met that still mean different things by entropy. That does not speak against the workshop; it shows where the field stands. That is exactly what day three was for: in working groups, participants were to sketch a roadmap for the field and work toward shared terms. We cannot report what came out of it, because we were not there. But the step is needed, because whether aging is more program or accumulated chance can only be settled once everyone knows what they are measuring.

So what we mainly brought home are better questions. When someone says an intervention lowers entropy: what exactly changed, in which tissue, and what does that mean for the body? When an aging clock moves: does it reflect a program or accumulated chance? You can ask these questions today, even before every field agrees on one shared measure.