The Clocks That Read

A true story in four parts, with one invented voice. By Jigs Shah. Part 3 (final text, v1, 29 Oct 2026).

Imagined image. A single sample tube catching early light, its thread wound about a clock with no numerals: the marks in our blood can be read, but no clock was set.

Part 3: The Clocks That Read

Told by the genes.

No clock was set

You would like a number. Most people would. Send a tube of blood away, and let something come back that says not how many birthdays you have had, but how old you really are.

We understand the wish, and we should be honest about what we are. There is no clock inside us that was wound at your conception and set to run down. An earlier story in this collection, The Long Slope, made that case: bodies grow more likely to die with age not because a clock was set at birth, but because repair was budgeted for a shorter, more dangerous life than most people now live [1]. It described a budget, a slope and chance. We will not tell that story again.

But the last part of this story ended on a question it could not leave alone. Over a life, our copies drift. Neurons gather new letters year by year; the cells of identical twins part ways a few letters at a time [2, 3]. If that drift is steady enough, can it be read? Can a cell’s text, or the way that text is opened, tell you how long it has been living?

The surprising answer is yes, more or less. There is no clock that was set. But there are records, and records can be read. What follows is the story of the people who learned to read them, of how well they read, and of where the reading stops being a measurement and starts being a promise.

A pact named for a king

Begin with three teenagers in Germany. Steve Horvath, his identical twin Markus and their friend Jörg Zimmermann met regularly to talk about mathematics, physics and philosophy, and they called their meetings the Gilgamesh project, after the Sumerian epic in which a king of Uruk searches for a plant that can restore youth [4]. At their last meeting, in 1989, they made a pact to spend their careers on science that might lengthen healthy human life. Steve said he would come at it through “mathematical modelling and gene networks” [4].

The way in, when it came, was sideways. In 2011, as a statistician at the University of California, Los Angeles, Horvath was asked to take part, with Markus, in a study of DNA methylation and sexual orientation. Markus is gay and Steve is straight; their DNA is identical. The study found no signal for orientation. Horvath then looked at the same data for age, and the signal was so strong that, in his words, “I fell off my chair” [5, 6].

To see what he saw, you need to know about a kind of mark we carry that is not a letter. Along our length sit small chemical tags called methyl groups. They do not change what we say. They change which of us is read: attached at particular places, they help turn genes on or off, and they can be kept as cells divide [7]. In Part 1 we called the cell a reader. Methyl marks are something like pencil notes in the margins of a book that tell the reader which pages to skip. Over a life, some of those notes are rubbed out and others are added, and not at random.

In 2013 Horvath published what became the best known of these readings. He gathered 8,000 samples from 82 datasets, covering 51 healthy tissues and cell types, and built a predictor from the methylation state of 353 places on our length, each a C followed by a G [8]. Fed a sample of almost any tissue, it returned an estimate of age. The estimate was close to zero for embryonic stem cells and for induced pluripotent stem cells, cells that had been coaxed back to an embryo-like state. It was strongly heritable. It worked, too, on chimpanzee tissue. And across 6,000 cancer samples from 20 cancer types, the tumours read as older than the people who carried them, by an average of 36 years [8].

The striking thing was not only that the reading worked. It was that it worked in so many places at once. A white blood cell may be a few days or weeks old, yet it carries the signature of the 50-year-old who made it, give or take a few years, and so does a cheek cell, or a piece of brain, or colon [4]. The marks did not record how old the cell was. They recorded, somehow, how long the body had been running.

Horvath offered an explanation as a proposal, not a finding: that methylation age “measures the cumulative effect of an epigenetic maintenance system” [8]. Keep that word proposal in mind. Much of what follows depends on it.

Imagined image. Two ways to tell age from one sample: the calendar’s clock, and a reading of marks at 353 places on our length, the basis of Horvath’s 2013 clock.

Generations of clocks

Horvath was not alone. In the same year, Gregory Hannum and colleagues published a clock built from the blood of 656 people aged 19 to 101 [9]. Both clocks were trained in the same way: given samples from people of known age, find the marks that best predict the number of birthdays. That design has a built-in limit, and the scientists who came next said so plainly: “the first generation of epigenetic biomarkers of aging were developed using chronological age as a surrogate for biological age” [10]. A clock trained to guess your birthday will, at best, guess your birthday. What people wanted to know was something else: whose body was wearing faster.

So the second generation was trained on outcomes rather than calendars. Morgan Levine’s PhenoAge, published in 2018, was built to track a “phenotypic age” drawn from clinical measures, and then shown to predict disease and death [10]. GrimAge, from Ake Lu, Horvath and colleagues in 2019, went further: it was assembled from methylation-based estimates of blood proteins and of smoking history in pack-years, and it was tuned to predict lifespan and healthspan [11].

The third generation asked a different question again. Not how old, but how fast. In Dunedin, New Zealand, a group of people born in 1972 and 1973 had been measured again and again across their lives. From repeated measures of 19 indicators of how their organ systems were working, researchers built a methylation measure called DunedinPACE, short for “Pace of Aging Calculated from the Epigenome” [12]. Where the first clocks were clock faces, this one is closer to a speedometer. It reads a rate rather than an age: how fast the body is changing, not how far it has come.

Generations of clocks: what each was trained to predict, and what it reads.

Each generation answered a fair criticism of the last. None of them is a clock in the sense of the wish we began with. They are statistical readings of marks, trained on large groups of people, and they are only as good as what they were trained on.

The same blood, twice

A good instrument gives the same answer twice. In 2022, a team led by Albert Higgins-Chen and Morgan Levine tested exactly that. They took samples that had been measured more than once, the same blood or saliva read repeatedly, and found that technical noise alone produced “deviations up to 9 years between replicates for six prominent epigenetic clocks” [13]. The same tube, read twice, could come back nearly a decade apart.

The fault lay less in the idea than in the arithmetic. Individual marks are measured imperfectly, and a clock that leans heavily on a few hundred of them inherits their wobble. The same team rebuilt six clocks from broader statistical summaries of many marks, called principal components, and their new versions brought “agreement between most replicates within 1.5 years” [13]. Better, but still not the precision a single number on a report seems to promise.

There is a second thing a fair reader should know. Many of these measures are also products. The paper describing the steadier clocks discloses that two of its authors built metrics licensed to Elysium Health through Yale University, a company in the ageing business [13]. Disclosures like this are a normal part of science, and they do not make the findings wrong. They do mean that the people who build the clocks and the people who sell them are sometimes the same people, and readers deserve to know that.

The harder test is whether a clock can register a change that someone deliberately made. The CALERIE trial offered one. In it, 220 adults without obesity were randomly assigned either to cut their calorie intake by 25 percent or to eat as usual, for two years [14]. When their stored blood was later read with the clocks, the restriction group showed a slower pace of ageing by DunedinPACE. But PhenoAge and GrimAge did not change significantly, and the authors described the treatment effects as small [14]. One careful trial; three clocks; two kinds of answer. The clocks do not always agree about the same people, because they were built to answer different questions.

It is worth pausing on what this means for the tube of blood in the wish we began with. A clock reading is not like a weight on a scale, which you can trust to the nearest few hundred grams. It is closer to a reading from a well-made survey: very good at describing a population, less sure about any one respondent, and best interpreted with its margin of error in view. A result that says you are three years “older” than your birthday may be telling you something about your body, or something about the day and the machine your sample was run on. Read once, it cannot easily say which.

A thermometer is not a fever

Here is the question beneath all the others. When the marks shift with age, are they causing the ageing, or merely recording it?

A thermometer rises with a fever. Breaking the thermometer does not cure anyone. If methylation is a thermometer, then pushing it back might change the reading without changing the body. If it is closer to a cause, then pushing it back might matter.

The evidence does not yet settle this, and honest people disagree. On one side sits a 2023 study from David Sinclair’s laboratory in which mice were engineered so that their DNA could be cut and faithfully repaired at chosen moments. The repeated repair, the authors reported, “advances aging at physiological, cognitive, and molecular levels”, including the methylation clock, and the changes could be reversed. They read their results as “consistent with the information theory of aging, which states that a loss of epigenetic information is a reversible cause of aging” [15]. That is a hypothesis with a mouse experiment behind it. Several of its authors, including Sinclair, disclose interests in companies developing rejuvenation treatments [15].

On the other side is the simple fact that correlation was all the first clocks were built to capture. A clock that tracks age well need not tell you why anyone ages. Horvath himself framed his explanation as a proposal [8]. For now, the safest sentence is the neutral one: the marks are a remarkably good record, and whether they are also a lever is an open question.

The marks at the ends

There is an older candidate for the body’s clock, and it deserves one fair paragraph. The caps at the ends of our strands, the telomeres, shorten as many cells divide; The Long Slope told their story, and the story of the cells that stop dividing [1]. What matters here is how well a telomere reading predicts anything. A review in Science in 2015 concluded that greater overall telomere attrition “predicts mortality and aging-related diseases” in people with inherited telomere disorders and in general human cohorts [16]. An earlier review had called the human evidence that telomere length is a biomarker of ageing “equivocal” [17]. And a large genetic study suggested that longer telomeres “increase risk for several cancers but reduce risk for some non-neoplastic diseases, including cardiovascular diseases” [18]. Put plainly: shorter telomeres go with higher death rates across large groups, but a telomere test is a poor predictor for any one person.

The letters that wear

Methyl marks are pencil notes. But Part 2 told of a second kind of drift, deeper and more permanent: changes in the letters themselves, gathered by single cells across a life [2]. If the notes can be read as a clock, what about the letters?

Part 2 gave a first taste of this. Single neurons from the brain’s prefrontal cortex gather new single-letter changes at roughly 23 a year, and those of the hippocampus at roughly 40, rising in a nearly straight line with age [2]. A neuron, once made, is with you for life, so its letters are a diary that is never thrown away. But a diary kept by one kind of cell in one species cannot tell you whether the writing has anything to do with how long a life lasts. For that, you would need to compare animals that live for very different lengths of time.

In 2022 a team led by Alex Cagan at the Wellcome Sanger Institute went looking across species. They isolated 208 intestinal crypts, small pockets in the lining of the gut each kept up by a few stem cells, from 56 individual animals of 16 species: black-and-white colobus monkey, cat, cow, dog, ferret, giraffe, harbour porpoise, horse, human, lion, mouse, naked mole-rat, rabbit, rat, ring-tailed lemur and tiger [19]. In every species with enough samples, letter changes piled up steadily with age.

The rates were wildly different. Across the 15 species whose ages were known, the yearly count of single-letter changes per genome ranged from 47 in humans to 796 in mice [19]. And the rate followed one thing above all. It showed “a strong inverse relationship with species lifespan, with no other life-history trait studied showing a comparable association” [19]. A giraffe and a naked mole-rat gather changes at nearly the same pace, 99 and 93 a year, and have similar lifespans, though the giraffe is around 23,000 times heavier [19].

Then came the figure that makes the study unforgettable. Lifespan across these species varied about 30-fold, and body mass about 40,000-fold. Yet by the end of a typical life, the burden of letter changes in a crypt varied only by “a factor of around 3”, with a mean of 3,206.4 substitutions per genome per crypt [19]. A mouse that lives a few years and a human who lives many decades arrive at the end with roughly similar numbers of changes. Short lives write fast; long lives write slowly; the total at the end is not so different.

It is tempting to say this proves that mutations cause ageing. The authors did not say that. They wrote that somatic mutation rates “may be a contributing factor in ageing”, and they listed other candidates, telomere attrition, epigenetic drift and failing protein upkeep among them, whose “causal roles and relative contributions remain debated” [19]. A tight pattern across 16 species is a powerful clue. It is not yet a mechanism.

Single-letter changes gathered each year in a gut crypt, from 47 in humans to 796 in mice, and the mean total by the end of a typical life across 16 species: 3,206.4, varying only by a factor of around 3. Data: Cagan et al. 2022 [19].

How much is in the letters?

If letters drift and marks shift, how much of the length of a life is written in what you inherited? Here the numbers seem to quarrel.

In 2018 J. Graham Ruby and colleagues, at Calico Life Sciences and the genealogy company Ancestry, analysed public family trees that included hundreds of millions of historical people. Their key move was to look not only at blood relatives but at in-laws, people who share a household and a life but no inherited letters. Spouses’ lifespans were correlated too, a sign that people tend to marry people like themselves in ways that affect lifespan. Once that was accounted for, they concluded that the heritability of human longevity for people born across the 1800s and early 1900s was “well below 10%”, lower than earlier estimates that had been “consistently low (∼15–30%)” [20].

Then, in January 2026, a team from the Weizmann Institute of Science and elsewhere, led by Uri Alon, published a paper in Science with a title that seemed to say the opposite: “Heritability of intrinsic human life span is about 50% when confounding factors are addressed” [21]. Twin studies, they noted, showed heritability “of only 20 to 25%”, and large pedigree studies suggested “as low as 6%”. Their argument was that these figures mix two kinds of death together. Some deaths come from outside, from accidents and infections; others come from within, from the body’s own decline. Correcting for the outside kind, with mathematical models and twins raised together and apart, they found that the heritability of lifespan due to intrinsic mortality “is above 50%” [21].

These are not quite two answers to one question. They are two rulers held against different things. One measures all deaths in historical families, where infection and accident took many lives that no inherited letter could have saved. The other tries to measure only the deaths that come from within. The Long Slope split mortality the same way, into chance and deterioration [1]. In a world full of outside deaths, the letters matter less to when you die; in a world where most outside deaths are prevented, they may matter more.

One more caution belongs here. Heritability is a statement about variation in a population, not about a person [22]. “About 50 percent” does not mean that half of your lifespan was written at your conception. It means that, in the populations studied and under their conditions, about half of the differences in intrinsic lifespan between people tracked differences in what they inherited. Change the conditions, and the number can change.

Imagined image. Two rulers for one question: counting all deaths in families of the 1800s–1900s, the heritability of lifespan reads well below 10%; counting only deaths from within, above 50%. The outside causes are what the second ruler leaves out.

Turning the hands back

The first clocks held a hint that became the most hopeful thread in this story. Horvath’s reading was close to zero in embryonic stem cells, and close to zero, too, in induced pluripotent stem cells [8]. Those were adult cells that had been made young again.

That had first been done in 2006. Kazutoshi Takahashi and Shinya Yamanaka found that four factors, Oct3/4, Sox2, c-Myc and Klf4, were enough to turn mouse skin cells into cells with the powers of embryonic stem cells [23]. It is one of the great surprises of modern biology: a cell’s identity, which seemed fixed, could be wound back. The 2012 Nobel Prize in Physiology or Medicine went jointly to John Gurdon and Shinya Yamanaka “for the discovery that mature cells can be reprogrammed to become pluripotent” [24]. Note the warning written into the same work: when the reprogrammed cells were placed in mice, they formed tumours [23]. A cell wound all the way back forgets what it was for.

So the next idea was to wind back only partly. In 2016 Alejandro Ocampo, Juan Carlos Izpisua Belmonte and colleagues switched the four factors on in short, repeated pulses in living mice. In a mouse model of premature ageing, this improved signs of ageing and prolonged lifespan; in older normal mice, it improved recovery from metabolic disease and muscle injury [25]. The lifespan result was in mice that age abnormally fast, not in normal mice. In 2020, Yuancheng Lu, David Sinclair and colleagues used three of the factors, leaving out the cancer-linked c-Myc, in the eyes of mice. The treatment restored youthful methylation patterns in retinal nerve cells and reversed vision loss in a mouse model of glaucoma and in aged mice [26].

This year the idea reached a person. In January 2026 the company Life Biosciences announced that the US Food and Drug Administration had cleared a first trial of its three-factor treatment, ER-100, in people with two diseases of the optic nerve, and in June it reported that the first participant had been dosed [27, 28]. On 8 October 2026 the company released early data: in three people with glaucoma, a single injection into the eye was well tolerated through 56 days, and two of the three showed preliminary improvement in visual-field tests [29]. These are a company’s interim results, from an open-label trial in three people. They are a beginning, not a verdict.

Imagined image. Turning the hands back: reprogramming restored vision in mice, and in 2026 a first trial began in people with diseases of the optic nerve.

What the readings promise, and what they do not

Get Jignesh Shah’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

The step from a reading to a remedy is where hope runs ahead of evidence, and it is worth drawing the line clearly.

What the readings promise, and what they do not. Bracketed numbers are the endnotes, linked below.

None of this makes the clocks useless. A reading that is noisy for one person can still be powerful across many. A trial that wants to know whether a treatment slows ageing cannot wait decades for its participants to die; it needs a measure it can read in two years. That is why scientists keep building better clocks, and why the CALERIE result mattered even though it was small [14].

What the clocks cannot read

Step back, and the shape of the story is clear. There was never a clock set at birth. There are records. Pencil notes in the margins shift as the decades pass, and with enough of them a statistician can guess the decades. Letters wear in every cell that copies itself, and across sixteen mammals the wear seems scaled to the length of a life. Some of what decides how long a body lasts is inherited, how much depending on what else is trying to kill it. And at least some of the notes can be rubbed out and rewritten, in a dish, in a mouse, and now, tentatively, in a human eye.

What the clocks cannot read is the person. They read a crowd of our copies, averaged, the way the laboratory in Part 2 read a crowd and called it you. Imagine a clock that reads 52 for a 50-year-old. It has said something about the marks in a tube of blood, read once, against a population. It has not said how that person’s life will go.

There is something fitting in this. For most of history, the only record of a life’s length was kept outside the body: in a parish register, a horoscope cast at birth, a date carved on stone. Now it turns out that the body keeps its own registers, several of them, written in different inks and at different speeds, none of them complete. The notes in the margins fade and are redrawn. The letters wear. The caps at the ends fray. No single register is the truth of a life; each is a partial account, kept by a different clerk, and they do not always agree. What a reader can do, with care, is lay them side by side and see where they point.

That is a humbler gift than the one the wish asked for. It is also a real one. A measure that can be read in two years, rather than waited on for forty, lets a treatment be tested within a single career. The clocks are not oracles. They are instruments, and like the first telescopes and the first microscopes they are rough, argued over and improving.

Jigs has his own view of where this leads, and it belongs to him, not to us. He thinks that medicine written for one person’s letters, the personalised mRNA and DNA treatments now being built, is worth its risks, and that worries about what it will cost should not be allowed to stop it. Others weigh the risks and the costs differently, and Part 4 will set out both sides.

Because the question has changed. For three parts we have been read: our letters by sequencers, our drift by single-cell surveys, our marks by clocks. Each time, the reading has been more precise and more personal than the last. If the records can be read, can they be rewritten? Not in a dish, not in a mouse, but in a person, on purpose, letter by letter?

That is the last part of the story. Part 4: The Editor’s Pen.

Endnotes

1. Shah, J. (2026). “The Long Slope.” Episode 2 of the companion series on death (thesis: “Not because a clock was set at birth, but because repair was budgeted for a shorter, more dangerous life than the one most of us now live”; Gompertz’s “chance” and “deterioration”; telomeres and senescence covered there). Unpublished draft, /workspace/death-project/series/episode-02-process-final.md
2. Lodato, M. A. et al. (2018). “Aging and neurodegeneration are associated with increased mutations in single human neurons.” Science 359:555–559 (161 single neurons in all: 119 from 15 neurologically normal individuals, 42 from 9 with Cockayne syndrome or xeroderma pigmentosum; somatic single-letter variants rise roughly linearly with age, ~23 per year in prefrontal cortex and ~40 per year in the hippocampal dentate gyrus; mutations in post-mitotic neurons “are unique to each cell, and thus can only be comprehensively assayed by comparing the genomes of single cells”). https://pmc.ncbi.nlm.nih.gov/articles/PMC5831169/
3. Jónsson, H. et al. (2021). “Differences between germline genomes of monozygotic twins.” Nature Genetics 53:27–34 (“monozygotic twins differ on average by 5.2 early developmental mutations”; about 15% of pairs carry a substantial number specific to one twin). https://pubmed.ncbi.nlm.nih.gov/33414551/
4. Nature news feature (2014). “Biomarkers and ageing: The clock-watcher.” Nature 508:168 (Horvath, identical twin Markus and Jörg Zimmermann formed “the Gilgamesh project”; final meeting in 1989 and pact; Steve “planned to use mathematical modelling and gene networks to understand how to extend life”; white blood cells “will carry the signature of the 50-year-old donor they came from, plus or minus a few years”). https://doi.org/10.1038/508168a
5. MIT Technology Review (15 Apr 2022). “Aging clocks aim to predict how long you’ll live” (2011 saliva study of sexual orientation with twin Markus; “Steve is straight and Markus is gay”; “I fell off my chair, because the signal was huge for aging”). https://www.technologyreview.com/2022/04/15/1050019/aging-clocks/
6. TIME (2026). “Can Aging Be Measured — and Eventually Reversed?” Interview with Steve Horvath (study led by Eric Vilain at UCLA; “Markus is gay and I’m straight”; “we didn’t find any signal for homosexuality”). https://time.com/collections/future-of-living/7357365/steve-horvath-longevity-aging/
7. National Human Genome Research Institute. “Methylation.” Talking Glossary of Genomic and Genetic Terms (“may be retained as cells divide”; methyl groups “turn a gene on or off”). https://www.genome.gov/genetics-glossary/Methylation
8. Horvath, S. (2013). “DNA methylation age of human tissues and cell types.” Genome Biology 14:R115 (8,000 samples from 82 datasets, 51 healthy tissues and cell types; close to zero for embryonic and induced pluripotent stem cells; highly heritable; applicable to chimpanzee tissue; 6,000 cancer samples, 20 cancer types, average age acceleration 36 years; 353 CpG sites; “I propose that DNA methylation age measures the cumulative effect of an epigenetic maintenance system”). https://pubmed.ncbi.nlm.nih.gov/24138928/
9. Hannum, G. et al. (2013). “Genome-wide methylation profiles reveal quantitative views of human aging rates.” Molecular Cell 49:359–367 (656 individuals aged 19 to 101). https://pubmed.ncbi.nlm.nih.gov/23177740/
10. Levine, M. E. et al. (2018). “An epigenetic biomarker of aging for lifespan and healthspan.” Aging 10:573–591 (“the first generation of epigenetic biomarkers of aging were developed using chronological age as a surrogate for biological age”; DNAm PhenoAge). https://pubmed.ncbi.nlm.nih.gov/29676998/
11. Lu, A. T. et al. (2019). “DNA methylation GrimAge strongly predicts lifespan and healthspan.” Aging 11:303–327 (built from DNAm surrogates of plasma proteins and an “estimator of smoking pack-years”). https://pubmed.ncbi.nlm.nih.gov/30669119/
12. Belsky, D. W. et al. (2022). “DunedinPACE, a DNA methylation biomarker of the pace of aging.” eLife 11:e73420 (Dunedin Study 1972–1973 birth cohort; “19 indicators of organ-system integrity across four time points spanning two decades”; “Pace of Aging Calculated from the Epigenome”). https://pubmed.ncbi.nlm.nih.gov/35029144/
13. Higgins-Chen, A. T. et al. (2022). “A computational solution for bolstering reliability of epigenetic clocks.” Nature Aging 2:644–661 (“deviations up to 9 years between replicates for six prominent epigenetic clocks”; principal-component versions “show agreement between most replicates within 1.5 years”; competing interests: metrics “licensed by Elysium Health through Yale University”). https://pubmed.ncbi.nlm.nih.gov/36277076/
14. Waziry, R. et al. (2023). “Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial.” Nature Aging 3:248–257 (n = 220 adults without obesity, 25% caloric restriction or ad libitum for 2 years; slowed DunedinPACE but “did not lead to significant changes” in PhenoAge and GrimAge; “Treatment effect sizes were small”). https://pubmed.ncbi.nlm.nih.gov/37118425/
15. Yang, J.-H. et al. (2023). “Loss of epigenetic information as a cause of mammalian aging.” Cell 186:305–326 (ICE mice; faithful DNA repair “advances aging at physiological, cognitive, and molecular levels”, reversible by OSK; “consistent with the information theory of aging, which states that a loss of epigenetic information is a reversible cause of aging”; D.A.S. discloses interests in Life Biosciences and others). https://pubmed.ncbi.nlm.nih.gov/36638792/
16. Blackburn, E. H., Epel, E. S. & Lin, J. (2015). “Human telomere biology: A contributory and interactive factor in aging, disease risks, and protection.” Science 350:1193–1198 (“Greater overall telomere attrition predicts mortality and aging-related diseases in inherited telomere syndrome patients, and also in general human cohorts”). https://pubmed.ncbi.nlm.nih.gov/26785477/
17. Mather, K. A. et al. (2011). “Is telomere length a biomarker of aging? A review.” Journals of Gerontology A 66:202–213 (“the evidence suggesting telomere length is a biomarker of aging in humans is equivocal”). https://pubmed.ncbi.nlm.nih.gov/21030466/
18. Telomeres Mendelian Randomization Collaboration; Haycock, P. C. et al. (2017). “Association Between Telomere Length and Risk of Cancer and Non-Neoplastic Diseases: A Mendelian Randomization Study.” JAMA Oncology 3:636–651 (“longer telomeres increase risk for several cancers but reduce risk for some non-neoplastic diseases, including cardiovascular diseases”). https://pubmed.ncbi.nlm.nih.gov/28241208/
19. Cagan, A. et al. (2022). “Somatic mutation rates scale with lifespan across mammals.” Nature 604:517–524 (“208 individual intestinal crypts from 56 individuals across 16 species”; 47 substitutions per year in humans to 796 in mice; giraffe 99 and naked mole-rat 93 per year, “around 23,000-fold” difference in body mass; “strong inverse relationship with species lifespan, with no other life-history trait studied showing a comparable association”; end-of-lifespan burden “varied only by a factor of around 3” despite ~30-fold lifespan and ~40,000-fold body-mass variation; mean “3,206.4 substitutions per genome per crypt, 95% confidence interval 2,683.9–3,728.9”; “may be a contributing factor in ageing”; causal roles of damage types “remain debated”). https://pmc.ncbi.nlm.nih.gov/articles/PMC9021023/
20. Ruby, J. G. et al. (2018). “Estimates of the Heritability of Human Longevity Are Substantially Inflated due to Assortative Mating.” Genetics 210:1109–1124 (Ancestry public trees “including hundreds of millions of historical persons”; in-law correlations; heritability for 1800s–early-1900s birth cohorts “well below 10%”; prior estimates “consistently low (∼15–30%)”). https://pmc.ncbi.nlm.nih.gov/articles/PMC6218226/
21. Shenhar, B. et al. (2026). “Heritability of intrinsic human life span is about 50% when confounding factors are addressed.” Science 391:504–510 (twin studies “only 20 to 25%”, pedigree studies “as low as 6%”; extrinsic mortality “such as accidents or infections”; twins raised together and apart; intrinsic heritability “above 50%”). https://pubmed.ncbi.nlm.nih.gov/41610249/
22. Downes, S. M. & Matthews, L. “Heritability.” Stanford Encyclopedia of Philosophy (heritability as a population measure; critics note values are mistakenly attributed “to individuals rather than populations”). https://plato.stanford.edu/entries/heritability/
23. Takahashi, K. & Yamanaka, S. (2006). “Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors.” Cell 126:663–676 (four factors “Oct3/4, Sox2, c-Myc, and Klf4”; transplantation “into nude mice resulted in tumors”). https://pubmed.ncbi.nlm.nih.gov/16904174/
24. The Nobel Prize in Physiology or Medicine 2012, awarded jointly to Sir John B. Gurdon and Shinya Yamanaka “for the discovery that mature cells can be reprogrammed to become pluripotent”. https://www.nobelprize.org/prizes/medicine/2012/summary/
25. Ocampo, A. et al. (2016). “In Vivo Amelioration of Age-Associated Hallmarks by Partial Reprogramming.” Cell 167:1719–1733 (short-term cyclic OSKM “prolongs lifespan in a mouse model of premature aging”; in older wild-type mice “improves recovery from metabolic disease and muscle injury”). https://pubmed.ncbi.nlm.nih.gov/27984723/
26. Lu, Y. et al. (2020). “Reprogramming to recover youthful epigenetic information and restore vision.” Nature 588:124–129 (OSK in mouse retinal ganglion cells “restores youthful DNA methylation patterns”, “reverses vision loss in a mouse model of glaucoma and in aged mice”). https://pubmed.ncbi.nlm.nih.gov/33268865/
27. Life Biosciences press release (28 Jan 2026). “Life Biosciences Announces FDA Clearance of IND Application for ER-100 in Optic Neuropathies” (OSK; open-angle glaucoma and NAION; NCT07290244). Company statement. https://www.lifebiosciences.com/life-biosciences-announces-fda-clearance-of-ind-application-for-er-100-in-optic-neuropathies/
28. Life Biosciences press release (9 Jun 2026). “Life Biosciences Announces First Patient Dosed in Phase 1 Trial of ER-100 for Optic Neuropathies.” Company statement; also reported by Nature news, 9 Jun 2026 (https://www.nature.com/articles/d41586-026-01836-7). https://www.lifebiosciences.com/life-biosciences-announces-first-patient-dosed-in-phase-1-trial-of-er-100-for-optic-neuropathies/
29. Life Biosciences press release (8 Oct 2026). “Life Biosciences Announces First-in-Human Data from Ongoing Phase 1 Trial Evaluating ER-100 in Optic Neuropathies” (three open-angle glaucoma participants; single intravitreal dose “well-tolerated through Day 56”; improvement in visual fields “in 2 of 3”; open-label). Company-reported interim data, not peer reviewed. https://www.lifebiosciences.com/life-biosciences-announces-first-in-human-data-from-ongoing-phase-1-trial-evaluating-er-100-in-optic-neuropathies/

Further reading

• Siddhartha Mukherjee, The Gene: An Intimate History (Scribner, 2016).
• Nessa Carey, The Epigenetics Revolution (Icon Books, 2011).
• Venki Ramakrishnan, Why We Die: The New Science of Aging and the Quest for Immortality (2024).
• Kathryn Paige Harden, The Genetic Lottery: Why DNA Matters for Social Equality (Princeton University Press, 2021).
• Horvath, S., “DNA methylation age of human tissues and cell types”, Genome Biology (2013). https://doi.org/10.1186/gb-2013-14-10-r115
• Belsky, D. W. et al., “DunedinPACE, a DNA methylation biomarker of the pace of aging”, eLife (2022). https://doi.org/10.7554/eLife.73420
• Higgins-Chen, A. T. et al., “A computational solution for bolstering reliability of epigenetic clocks”, Nature Aging (2022). https://doi.org/10.1038/s43587-022-00248-2
• Cagan, A. et al., “Somatic mutation rates scale with lifespan across mammals”, Nature (2022). https://doi.org/10.1038/s41586-022-04618-z
• Ruby, J. G. et al., “Estimates of the Heritability of Human Longevity Are Substantially Inflated due to Assortative Mating”, Genetics (2018). https://doi.org/10.1534/genetics.118.301613
• Shenhar, B. et al., “Heritability of intrinsic human life span is about 50% when confounding factors are addressed”, Science (2026). https://doi.org/10.1126/science.adz1187
• Ocampo, A. et al., “In Vivo Amelioration of Age-Associated Hallmarks by Partial Reprogramming”, Cell (2016). https://doi.org/10.1016/j.cell.2016.11.052

Image credits

1. gen-p3-cover.jpg: Illustration made for this series. AI-generated; an imagined scene, not a photograph or historical record.
2. gen-p3-clocks.jpg: Illustration made for this series. AI-generated; an imagined scene, not a photograph or historical record. Label added in code.
3. gen-p3-table-clocks.jpg: Table made for this series from the sources cited in the text.
4. gen-p3-cagan.jpg: Diagram made for this series, drawn in code from published figures (Cagan et al., Nature, 2022; endnote 19).
5. gen-p3-rulers.jpg: Illustration made for this series. AI-generated; an imagined scene, not a photograph or historical record. Labels added in code.
6. gen-p3-eye.jpg: Illustration made for this series. AI-generated; an imagined scene, not a photograph or historical record. Label added in code.
7. gen-p3-table-claims.jpg: Table made for this series from the sources cited in the text.

Image usage note

Every image in this document is an illustration made for this series. All are AI-generated and imagined; none is a photograph, a historical image or a scientific figure.