Epigenetic alterations
If your DNA is the hardware, the epigenome is the software — the layer of chemical switches that decides which genes are on or off in each cell, and it drifts out of tune as you age. This matters because, unlike DNA mutations, these marks are reversible, they are what "epigenetic clocks" read to estimate your biological age, and in trials a few ordinary measures, mainly eating less, have shifted them slightly. The honest caveat: the clocks are the most developed aging biomarkers we have but the consumer versions are noisy, and the dramatic "age reversal" results come from small studies and from mice.
Epigenetic alteration is the third of the twelve hallmarks of aging, and the one that best explains why behaviour matters so much. Your DNA sequence is essentially fixed for life, yet inherited gene variants explain only a minority of the variation in how long people live — about 20–25% in twin studies, and well under 10% in a family-tree study of people born in the 1800s and early 1900s once it allowed for people tending to marry partners who share lifespan-shaping traits and circumstances.[1] The rest is shaped by environment, diet, lifestyle and chance. That is a statement about spread across a population, not about any one person's lifespan. The split is contested: a 2026 reanalysis argues these estimates are deflated by deaths from accidents and infections, and puts the heritability of intrinsic ageing above 50% once those are stripped out.[2] The argument for behaviour does not rest on the exact figure. Those outside influences reach the genome through the epigenome: a layer of reversible chemical marks sitting on top of the DNA that controls which genes are switched on without changing the underlying code. As we age, that control layer degrades — cells lose their precise identities, and tissues drift away from the gene-activity patterns of youth.[3]
How confident should you be
Descriptive — Strong. Measurement — Strong for comparing groups, Weak for individuals. Intervention — Moderate at best.
That the epigenome drifts with age, and that the drift is reversible in principle, is well established. That methylation-based clocks track health and mortality across populations is Strong — they are the most developed aging biomarkers available, though not the best-performing on every test. That your clock reading means anything is Weak: the clocks disagree with one another, the older ones return different ages when the same blood sample is measured twice, and even the steadier ones wobble when the same person is re-tested within days. Everything aimed at moving the clock deliberately is Weak, with one Moderate exception: calorie restriction, tested in a randomised trial, though the methylation analysis within it was not pre-specified. A few other randomised trials, of omega-3 and a multivitamin, found smaller effects in analyses added after the fact or run as side studies.
The software of the cell
Three interlocking systems make up the epigenome, and all three slip with age.[4]
DNA methylation is the most studied. A small chemical tag (a methyl group) is attached to the DNA at specific spots, usually where a cytosine "letter" sits next to a guanine — a so-called CpG site. A cluster of these tags on a gene's control region typically switches that gene off. Methylation is what keeps a liver cell behaving like a liver cell and a skin cell like a skin cell, by silencing the genes each doesn't need.
Histone modifications work on the spools. DNA is wound around protein spools called histones, and chemical tags on those spools change how tightly the DNA is packed. Adding an acetyl tag (acetylation) loosens the packing and opens genes for reading; removing it tightens the packing and shuts them. The enzymes that add and remove these tags — and the ones that physically slide the spools along the DNA — are a major control system, and several of them are druggable.
Non-coding RNAs are a third layer: RNA molecules that don't code for proteins but fine-tune which genes are expressed.
The crucial property of all three is that they are reversible. A mutation is a permanent typo in the code; an epigenetic mark is a pencil annotation that can be erased and rewritten. That reversibility is exactly what makes the epigenome a target for intervention — and what separates this hallmark from genomic instability, where the damage is to the code itself.
How the aging epigenome drifts
Strong — the two-directional drift is one of the best-replicated findings in the field.
Aging doesn't simply add or remove marks uniformly; it scrambles them in a characteristic, two-directional way.[5]
Global loss of methylation. Across the genome as a whole, methylation tags are gradually lost — especially from the vast stretches of repetitive DNA that are normally kept silenced. Among those sequences are "jumping genes" (transposable elements): ancient stretches of parasitic DNA, some of them relics of old viral infections, that can reactivate when their methylation silencing erodes, adding to genetic damage and to sterile inflammation — inflammation with no infection behind it.[6] Most of the direct evidence for the inflammation link comes from senescent cells grown in the lab and from old mice, in which a drug that blocks the jumping genes' copying enzyme (reverse transcriptase) calmed age-related inflammation; it has not been shown in people.[7] This is one of the molecular threads tying epigenetic drift to inflammaging.
Local gain of methylation. At the same time, methylation accumulates on the control regions of specific important genes — including tumour-suppressor genes and genes governing cell identity and repair — switching them off when they're needed.[8] The combination is the worst of both worlds: the genes that should be quiet get noisy, and the genes that should be active get silenced.
A parallel structural change is the gradual unravelling of heterochromatin — the densely packed, switched-off regions of the genome laid down early in development. The "heterochromatin loss" model of aging holds that these silent domains progressively break down, exposing genes that should stay dormant.[9] The same machinery is destabilised in the accelerated-aging disease Hutchinson-Gilford progeria, where a faulty nuclear-envelope protein loosens heterochromatin's anchoring — one of the links between this hallmark and the progeroid syndromes discussed under genomic instability.
The enzymes you can nudge: histones and sirtuins
Weak / preclinical — striking results in mice, nothing established in people.
Because the marks are written and erased by enzymes, those enzymes are obvious drug targets, and animal work is encouraging if still early. Blocking the enzymes that remove acetyl tags — histone deacetylases, or HDACs — restored youthful chromatin and prevented age-related memory loss in mice; conversely, switching off a specific tag-adding enzyme called KAT7 relieved cellular senescence and extended lifespan in mice.[10] These are striking proofs of principle in animals, not human therapies.
The best-known of these enzymes are the sirtuins, a family of "deacetylases" that strip acetyl tags off histones and other proteins. Their importance for longevity is that they only work when fuelled by nicotinamide adenine dinucleotide (NAD⁺), a molecule whose levels fall with age — which makes sirtuins a sensor that ties your metabolic state to your chromatin. Specific sirtuins guard genomic stability and mitochondrial function, and SIRT6-deficient mice in particular age faster.[11] But the popular leap from "sirtuins matter" to "sirtuin-boosting supplements extend life" is not supported. The early worm and fly lifespan results largely disappeared once genetic background was properly controlled.[12] In mice the picture depends on which sirtuin and where: extra SIRT1 throughout the body did not lengthen life, while extra SIRT6 lengthened median lifespan by about 27% in males and 15% in females — though both SIRT6 results come from a single laboratory, and its earlier study, in a different mouse strain, saw the effect in males alone.[13] Lifelong genetic overexpression is not the same thing as taking a pill. This is the mechanistic backdrop to the NAD⁺ precursors and resveratrol covered under geroprotectors — real biology, thin human longevity evidence.
Epigenetic clocks: the most developed aging biomarkers we have
Strong as a research tool for comparing groups; Weak for any individual reading. They are the most developed human aging biomarkers, but "most developed" is not "best": in one US cohort of adults over 50, a score built from seven inflammation blood measures predicted four-year mortality slightly better than the six clocks it was compared with,[14] while in about 1,100 Berliners aged 60 to 80 followed for about seven years, the pace-of-aging clock DunedinPACE was the strongest predictor of death among 14 proposed aging markers. The inflammation marker interleukin-6 (IL-6) and grip strength also predicted death, though less strongly, and the inflammation marker C-reactive protein (CRP) showed no link.[15] Which marker comes out ahead may depend on the group studied and the clock used. See the hallmarks hub.
The single most consequential product of this field is the epigenetic clock — an algorithm that reads methylation at a few hundred CpG sites and returns an estimate of biological age that can differ from your age in years. The clocks have evolved through generations, each better aimed at health rather than just calendar age.[16]
- First generation (Horvath, Hannum) was trained to predict chronological age. Accurate at that, but only loosely tied to health.
- Second generation (PhenoAge, GrimAge) was trained against clinical markers and mortality, and these are the ones that actually track health: across nine clocks estimated from a single large cohort, the first- and fourth-generation versions were not as closely tied to socioeconomic status or health outcomes as the second- and third-generation ones.[17]
- Third generation (DunedinPACE) doesn't report an age at all — it reports a rate, the pace of biological aging per calendar year. It was built for intervention trials and gives near-identical results when the same sample is re-run, but in one study its readings in the same person shifted within hours after a meal or a stress test, so a single person's change on it needs caution.[18]
- Fourth and fifth generation (causality-enriched and organ-system clocks) try to isolate the methylation changes that actually drive damage rather than merely correlate with age — the same causal logic behind the DamAge and AdaptAge metrics discussed under hallmarks of aging.
The largest head-to-head test so far, in nearly 19,000 Scottish adults followed for ten years, agrees with the generational ranking above: the second- and third-generation clocks predicted later disease and death clearly better than the first-generation ones. But adding a clock to a model of standard risk factors improved prediction by more than one percentage point in fewer than one in five of the clock-disease links it found.[19]
The clocks have genuine biological grip: higher body weight and metabolic dysfunction show up as accelerated clocks, and an accelerated clock can appear in the blood of women years before a breast-cancer diagnosis.[20] Lower socioeconomic status does too, most clearly on the second- and third-generation clocks.[21] In a large German aging cohort, the pace-of-aging and organ-system clocks tracked cognitive decline most consistently — a result so far reported only in a preprint that has not been peer-reviewed.[22]
But can you trust your own clock result?
Weak — the consumer tests are not reliable enough to act on, and Caution applies to what is sold on top of them.
The leap from "powerful research tool" to "reliable personal test" is where the field gets oversold, and the caveats are the same ones that sink consumer telomere testing. Start with the most awkward one: the clocks do not agree with each other. Estimated side by side from a single cohort, nine widely used clocks read different sets of methylation sites and do not correlate strongly among themselves.[23] There is no single quantity called "your epigenetic age" that they are all approximating. Reliability is the deeper problem, and it comes in two layers. The first is lab noise: when the same blood sample was measured twice, the older clocks (Horvath, Hannum, PhenoAge) gave ages a median of about one to two and a half years apart, and in the worst case nearly nine, against a real between-person spread of only three to five years; GrimAge and re-engineered versions of the clocks were far steadier.[24] That noise comes from the individual methylation probes the clocks read.[25] Newer clocks have largely fixed that layer. The second layer has not been fixed: when the same person gives blood again within hours or days, before and after a meal or a stress test, most clocks show only low-to-moderate stability, and that short-term wobble is large enough to make a change in one person's result hard to interpret.[26]
- Sample type matters. Blood is the gold standard, because it's a well-characterised mix of immune cells. Many consumer kits use saliva or cheek swabs, which contain cheek-lining cells with entirely different methylation. When researchers ran widely used clocks on mouth (saliva and cheek-swab) and blood samples from 83 people aged 9 to 70, the mouth samples gave ages that differed from the same person's blood by close to 30 years on average on some clocks, and the two still matched poorly even after allowing for differences in cell mix; only one clock, built for skin and blood, agreed reasonably well across tissues.[27]
- Platform matters. Results can shift with the measurement technology, and providers' numbers are not directly comparable.
- No regulatory validation. No epigenetic clock is accepted as a clinical endpoint, and regulators such as the US Food and Drug Administration (FDA) and the European Medicines Agency have not yet issued any guidelines on how biomarkers of aging, clocks included, should be developed, measured or validated.[28] A survey of experts in the field listed, among its core recommendations, proving that these markers are reliable and respond to change within a single person.[29] Companies meanwhile often pair a proprietary, unvalidated algorithm with generic lifestyle advice.
So a single consumer clock reading is not something to act on. The clocks earn their keep in research — comparing groups, tracking interventions across many people — far more than in telling one individual their "true age."
What actually moves the clock
Weak overall, Moderate for calorie restriction. A handful of randomised trials with small effects, their clock analyses added after the fact or run as side studies, one pilot with 43 participants, and a set of observational associations.
The interventions with the best evidence for slowing epigenetic aging are, once again, the familiar ones — and the strongest single piece of evidence is a real randomised trial.
Caloric restriction has the best evidence — with one caveat worth stating up front. The CALERIE trial randomised 220 healthy adults without obesity to a 25% calorie-restriction target or an ad-libitum diet for two years, and the restricted group significantly slowed their pace of aging on the DunedinPACE clock. The caveat is that the methylation analysis was performed post hoc — it was not what the trial was designed to test — so this is randomised data analysed after the fact rather than a pre-specified result. Adherence fell well short of the target, at roughly 12% average restriction rather than 25%. The slowing was small: about 2–3% slower pace of aging in these adults, who were aged 21–50. From earlier observational studies, the authors estimate this could mean up to 10–15% lower risk of death. That is a projection, not something the trial measured, and two of the authors are named inventors on the clock that moved, which is licensed to a company.[30] Notably, the effect appeared on the pace-of-aging clock and not at all on the static clocks, PhenoAge and GrimAge — a clean null there, and a reminder that "which clock" changes the answer.
Diet supplies the raw material for methylation. The methyl tags themselves are built from nutrients in "one-carbon metabolism" — folate, vitamin B12, choline, and methionine — so a diet rich in these methyl donors is the biochemical rationale behind the dietary protocols aimed at the epigenome. A small eight-week pilot trial randomised 43 healthy men aged 50–72 (38 were analysed). It combined a methyl-donor-rich diet with sleep, exercise and relaxation guidance plus probiotic and plant-compound supplements. On an older, first-generation clock measured in saliva, the men on the programme ended about two years younger than they started, while the untreated comparison group drifted about a year older. That drift was not statistically significant, but it accounts for roughly a third of the headline three-year gap between the groups and hints at how noisy a single clock reading is. The study was funded by a supplement company, and two of its authors use the programme in their clinical practice, earn money from related educational products and are named on a related patent application. It needs replication before it means much.[31] Several plant compounds in that same protocol also nudge the relevant enzymes — broccoli-derived sulforaphane and dietary polyphenols modulate the acetyl-removing HDACs and the methylation-adding DNA methyltransferases (DNMTs) — though these are mechanistic signals, not proven longevity interventions.[32]
Exercise, sleep, and stress. Exercise is the lever most often said to slow the clocks, and the randomised evidence for that is thin and conflicting, which is worth saying plainly. A large three-year trial in 777 generally healthy, active Swiss adults aged 70 and over compared a home strength programme three times a week with a stretching programme, and exercise on its own did not change any of the four modern clocks it measured; it showed up only as a small add-on to omega-3 and vitamin D on one clock.[33] A smaller trial in 124 sedentary breast-cancer survivors found that a year of exercise slowed a mortality-linked clock (GrimAge2) and a fitness-based clock compared with a health-and-wellness programme; the clocks were measured in only part of the trial, and the study summary does not say how large the effect was.[34] Exercise earns its place on this site through its effects on muscle, heart and survival, not through the clocks. Sleep is less simple than usually reported. In the Young Finns cohort the robust and actionable finding is that symptoms of obstructive sleep apnoea track accelerated epigenetic ageing independently of other health factors — see sleep-disordered breathing. Shift work did not produce a clean dose-response: its effect showed on one clock in one model only, and among people with little shift-work history, insomnia and sleep deprivation were associated with more acceleration than in long-term shift workers, which the authors read as possible adaptation.[35] Acute sleep loss does shift methylation on the core clock genes.[36] Chronic psychological stress is widely held to accelerate the clocks, though the mechanism is not established here. The meditation evidence is narrower than it is usually reported: there was no overall difference between long-term meditators and controls, and the protection appeared only in those over about 52, alongside a small gradient of roughly a quarter of a year less epigenetic age acceleration per additional year of practice.[37] Such studies also cannot rule out that people who meditate for decades differ in other ways. A multidomain lifestyle programme in frail older adults — supervised exercise three days a week plus nutritional support — improved physical function and one of five epigenetic clocks in a randomised trial. It is very small: 47 people randomised, with methylation measured in only 24 of them, eight of those controls, after the pandemic disrupted recruitment. It reinforces the general pattern rather than settling anything — the package moved the needle, and this trial cannot say which part did the work.[38]
Two large supplement trials add small signals. In the same three-year trial of Swiss adults aged 70 and over described above, a daily 1 g omega-3 supplement slowed three of the four clocks: about 1% on the pace-of-aging clock and roughly three to four months on two others, while vitamin D on its own changed none of them. Like the calorie-restriction result, this was an after-the-fact analysis.[39] In a planned side study of 958 older Americans, a daily multivitamin slowed two second-generation clocks by roughly a tenth to a fifth of a year per year over two years, and a cocoa-flavanol extract moved none of the five clocks tested. The authors call the effects small and say more studies are needed to know whether they matter clinically.[40] A weight-loss drug has also slowed several clocks, in a small after-the-fact analysis of a randomised trial in adults with HIV-associated fat gain — see semaglutide and the aging clock.
The reprogramming frontier
Weak / preclinical, and Caution on the safety question — the same factors that rejuvenate cells cause tumours if left on.
The most radical idea in the field is partial reprogramming: briefly switching on the four "Yamanaka factors" — the same genes that can revert an adult cell all the way back to a stem-cell state — but only transiently, so that aged epigenetic marks are wiped clean while the cell keeps its identity. In mice, this approach restored youthful gene-activity patterns to nerve cells in the eye and reversed age-related vision loss.[41] Applied long-term and intermittently to normally aging mice, it rejuvenated skin and kidney, reversed the epigenetic clock, and damped genes for inflammation and senescence — with the authors concluding such regimens can be designed to be safe.[42] In one company-run study, a gene therapy carrying three of the factors was given to male mice already near the end of a normal mouse lifespan (124 weeks old). It about doubled their remaining lifespan, from about 9 to about 18 weeks, and made them less frail. The study used only males, did not report its group sizes in the main text and had no comparison group given an empty virus, so it needs repeating.[43]
The catch is severe: the same factors that rejuvenate cells can, if switched on too long or too strongly, push cells all the way back to a stem-cell-like state. In mice, switching on all four factors produced tumours called teratomas in several organs;[44] even a short burst followed by switching off left behind cancer-like growths, including kidney tumours,[45] and keeping the factors on continuously killed the animals within days.[46] Making the reset brief and tightly controlled is the central safety challenge. One line of work tries to replace the genetic factors with drug-like chemical cocktails. A screen found six combinations that made aged cells in a dish look younger by their gene-activity patterns within a week; the study was in cells only, and the authors say the cocktails must be tested for safety in animals before any human use.[47] The first human test has only just begun. In 2026 a small early-stage safety trial started giving a reprogramming gene therapy, injected into the eye, to people with glaucoma or a type of optic-nerve stroke; results are not expected before 2027.[48] It targets a specific eye disease, not ageing itself, and none of this is near ready for healthy people. This is a frontier worth watching, not a therapy to seek out.
What this does and doesn't tell you
What it tells you: epigenetic alterations are the hallmark that makes the case for behaviour — a large share of the variation in how we age is non-genetic and runs through reversible marks, though how large is genuinely contested and this site does not pick a number. The epigenetic clocks built on those marks are the most developed human biomarkers of aging, the calorie-restriction evidence (CALERIE) comes from a genuine randomised trial, albeit from a post hoc analysis within it, and the things shown in randomised trials to nudge the clocks are mostly dietary: eating less, and in older adults omega-3 and a multivitamin, all with small effects. For exercise the trials disagree — a large one in healthy older adults found nothing, a smaller one in cancer survivors found a benefit — and for sleep and stress there are no such trials yet.
What it doesn't tell you: that a consumer clock reading is precise enough to act on (it usually isn't), that any supplement reliably reverses your epigenetic age, or that reprogramming is anywhere near safe for people. The dramatic numbers — "three years younger in eight weeks," "vision restored" — come from a single small, supplement-company-funded pilot whose headline gap was partly drift in the untreated group, and from mice. The durable message is the unglamorous one: the epigenome is tunable in principle, the knobs proven in trials so far turn it only slightly, and the ordinary habits earn their place through their effect on health, not on a clock reading.
Further reading
- Dhar P, et al. Epigenetic alterations — the silent indicator for early aging and age-associated health-risks. Aging Med (Milton) 2022.[49]
- An Y, et al. Epigenetic regulation of aging and its rejuvenation. MedComm 2025.[50]
- Tsurumi A, Li WX. Global heterochromatin loss: a unifying theory of aging? Epigenetics 2012.[51]
- Lemus AJJ, et al. The interplay of epigenetic remodelling and transposon-mediated genomic instability in ageing and longevity. Open Biology 2026.[52]
- Crimmins EM, et al. Generations of epigenetic clocks and their links to socioeconomic status in the Health and Retirement Study. Epigenomics 2024.[53]
- Waziry R, et al. Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial. Nat Aging 2023 — post hoc analysis of a randomised trial.[54]
- Fitzgerald KN, et al. Potential reversal of epigenetic age using a diet and lifestyle intervention: a pilot randomized clinical trial. Aging (Albany NY) 2021.[55]
- Olaso-Gonzalez G, et al. A multidomain lifestyle intervention is associated with improved functional trajectories and favorable changes in epigenetic aging markers in frail older adults: a randomized controlled trial. Aging Cell 2026.[56]
- Autio I, et al. Sleep disturbances, shift work, and epigenetic ageing in working-age adults: findings from the Young Finns study. Clin Epigenetics 2025.[57]
- Cedernaes J, et al. Acute sleep loss induces tissue-specific epigenetic and transcriptional alterations to circadian clock genes in men. J Clin Endocrinol Metab 2015.[58]
- Browder KC, et al. In vivo partial reprogramming alters age-associated molecular changes during physiological aging in mice. Nat Aging 2022.[59]
- Lu Y, et al. Reprogramming to recover youthful epigenetic information and restore vision. Nature 2020.[60]
- Macip CC, et al. Gene therapy-mediated partial reprogramming extends lifespan and reverses age-related changes in aged mice. Cell Reprogram 2024.[61]
- Yang JH, et al. Chemically induced reprogramming to reverse cellular aging. Aging (Albany NY) 2023.[62]
- Chaix R, et al. Epigenetic clock analysis in long-term meditators. Psychoneuroendocrinology 2017.[63]
- Higgins-Chen AT, et al. A computational solution for bolstering reliability of epigenetic clocks: implications for clinical trials and longitudinal tracking. Nat Aging 2022.[64]
- Sugden K, et al. Patterns of reliability: assessing the reproducibility and integrity of DNA methylation measurement. Patterns (N Y) 2020.[65]
- Ruby JG, et al. Estimates of the Heritability of Human Longevity Are Substantially Inflated due to Assortative Mating. Genetics 2018.[66]
- Shenhar B, et al. Heritability of intrinsic human life span is about 50% when confounding factors are addressed. Science 2026.[67]
- De Cecco M, et al. L1 drives IFN in senescent cells and promotes age-associated inflammation. Nature 2019.[68]
- Burnett C, et al. Absence of effects of Sir2 overexpression on lifespan in C. elegans and Drosophila. Nature 2011.[69]
- Roichman A, et al. Restoration of energy homeostasis by SIRT6 extends healthy lifespan. Nat Commun 2021.[70]
- Vetter VM, et al. Comparing fourteen consensus biomarkers of aging: epigenetic pace of aging as the strongest predictor of mortality in BASE-II. Biomark Res 2026.[71]
- Mavrommatis C, et al. An unbiased comparison of 14 epigenetic clocks in relation to 174 incident disease outcomes. Nat Commun 2025.[72]
- Sehgal R, et al. Biological Versus Technical Reliability of Epigenetic Clocks and Implications for Disease Prognosis and Intervention Response. Aging Cell 2026.[73]
- Apsley AT, et al. Cross-tissue comparison of epigenetic aging clocks in humans. Aging Cell 2025.[74]
- Moqri M, et al. Validation of biomarkers of aging. Nat Med 2024.[75]
- Herzog CMS, et al. Challenges and recommendations for the translation of biomarkers of aging. Nat Aging 2024.[76]
- Bischoff-Ferrari HA, et al. Individual and additive effects of vitamin D, omega-3 and exercise on DNA methylation clocks of biological aging in older adults from the DO-HEALTH trial. Nat Aging 2025.[77]
- Carlin D, et al. Exercise slows biological aging in breast cancer survivors: secondary analysis of a randomized trial. J Natl Cancer Inst 2026.[78]
- Li S, et al. Effects of daily multivitamin-multimineral and cocoa extract supplementation on epigenetic aging clocks in the COSMOS randomized clinical trial. Nat Med 2026.[79]
- Abad M, et al. Reprogramming in vivo produces teratomas and iPS cells with totipotency features. Nature 2013.[80]
- Ohnishi K, et al. Premature termination of reprogramming in vivo leads to cancer development through altered epigenetic regulation. Cell 2014.[81]
- Ocampo A, et al. In Vivo Amelioration of Age-Associated Hallmarks by Partial Reprogramming. Cell 2016.[82]