Can You Tell a Person’s Age From Their DNA?

DNA methylation patterns can estimate a person’s chronological age with striking accuracy, often within a few years. The breakthrough behind this came from the discovery that chemical tags on DNA shift in predictable ways as we get older, forming what researchers call “epigenetic clocks.” These clocks now outperform every other DNA-based aging marker, and they are already being used in forensic investigations and medical research. But the gap between your DNA-predicted age and your actual birthday can itself be revealing, sometimes more so than the estimate itself.

How DNA Records the Passage of Time

Throughout your life, small chemical groups called methyl tags attach to or detach from specific spots on your DNA. These changes do not alter your genetic code; they alter how genes are read. Hundreds of these spots shift in a consistent, age-related direction across a person’s lifetime. By measuring the methylation status at a carefully chosen set of these locations, researchers can build a statistical model that predicts how old someone is.

The most influential of these models was developed using roughly 8,000 samples from 82 datasets, spanning 51 different tissue and cell types. It identified 353 specific sites on the genome that together function as a remarkably reliable aging clock. The predicted age starts near zero for embryonic stem cells and tracks closely with chronological age across childhood, adulthood, and old age.¹ This multi-tissue clock set the standard for the field and demonstrated that aging leaves a consistent molecular signature regardless of whether the DNA comes from blood, brain, liver, or skin.

How Accurate Is the Estimate?

Accuracy depends on which clock is used, which tissue is sampled, and the age range of the person being tested. Early models using just a handful of methylation sites achieved an average error of about five years.² The 353-site clock brought that down to about 3.6 years on average in test data.³ More recent forensic models, built with machine-learning algorithms and optimized sets of methylation markers, have pushed the average error to around 3.3 years, with predictions for people under 55 achieving errors as low as 2.6 years.⁴

The frontier is even sharper. A 2025 study using deep learning to analyze methylation patterns at the level of individual DNA molecules reported a median error of just 1.36 years in people under 50.⁵ That is an extraordinary level of precision for a biological measurement, though it has not yet been widely replicated in independent forensic settings. The trend is clear: as the analytical tools improve, the window of uncertainty keeps narrowing.

Why Telomere Length Does Not Work Well

Before methylation clocks took off, telomere length was the most talked-about DNA marker of aging. Telomeres are the protective caps at the ends of chromosomes, and they do shorten as cells divide over a lifetime. A large meta-analysis covering more than 743,000 individuals confirmed a negative correlation between telomere length and age, with a median shortening rate of about 23 base pairs per year in cross-sectional studies.⁶ The problem is that the variation between individuals is enormous. Two people who are exactly the same age can have wildly different telomere lengths.

A study that directly tested telomere length as an age predictor found the statistical relationship was far too weak for practical use, with an R-squared value of roughly 0.04. Even after splitting samples by sex or ethnicity, accuracy remained dismal. People of the same age showed such different telomere lengths that any prediction based on them was essentially a guess.⁷ Telomere length is influenced by genetics, stress, chronic illness, and lifestyle in ways that swamp the age signal. It tells you something about cellular wear and tear, but it cannot reliably tell you when someone was born.

The Tissue You Test Matters

One underappreciated wrinkle is that DNA methylation age can differ depending on where in the body the sample comes from. Blood, saliva, and cheek swabs generally produce consistent results, with average errors in the range of three to five years.⁸ Semen, however, gives significantly less accurate results, with errors of seven to eight years, likely because sperm cells have distinct methylation patterns shaped by reproductive biology rather than chronological aging.

A multi-tissue model built specifically for forensic use, trained on blood, saliva, and buccal swab samples, achieved an average error of about 3.9 years across all three. When tested on a real-world forensic sample type, chewed gum from 25 people, the error dropped to 2.8 years.⁹ The practical value here is obvious: at a crime scene, investigators may not know what kind of biological material they are working with. A clock that performs well regardless of tissue type is far more useful than one tuned for blood alone.

Even within blood, the type of preparation matters. Saliva-derived samples tend to show DNA methylation ages several years older than blood-derived samples from the same person, which means mixing tissue types without adjustment can introduce systematic bias into results.¹⁰

The Difference Between Chronological and Biological Age

If the only thing epigenetic clocks did was estimate your birthday, they would be a neat trick but not much more. Their real power is in the gap between your predicted methylation age and your actual age. When your DNA says you are older than your birthday says, researchers call that “age acceleration,” and it is linked to worse health outcomes.

First-generation clocks were designed purely to match chronological age as closely as possible. Second-generation clocks took a different approach, training on markers associated with disease risk and mortality rather than birth certificates. Third-generation clocks go further still, incorporating the pace at which individuals age over time, measured by tracking health decline across years, into a single blood-based snapshot.¹¹ These newer clocks are better at predicting who will develop age-related disease, even among people of identical chronological age. In other words, the question has shifted from “how old are you?” to “how fast are you aging?”

What Makes Your DNA Age Faster or Slower

Smoking is one of the strongest accelerators of epigenetic aging. It consistently shows up across studies as a driver of age acceleration, whether measured by first- or second-generation clocks. Higher body mass index and waist-to-hip ratio also push DNA age upward, as do markers of poor metabolic health like elevated blood sugar and uric acid.¹²

Education and socioeconomic factors play a role too. Higher education has been associated with a younger methylation age relative to chronological age, while smoking, alcohol consumption, and higher BMI all push it in the opposite direction.¹³ These effects are not just cosmetic statistical associations. A study of coronary heart disease risk found that roughly 10% of the elevated heart disease risk from smoking and about 18% of the risk associated with certain dietary patterns were mediated through methylation aging.¹⁴ The methylation changes are not just recording damage; they appear to be part of the pathway through which lifestyle factors lead to disease.

Predicting Age in Children

Most epigenetic clocks were built using adult samples, and they struggle with children. Methylation changes during childhood are more rapid and dynamic than in adulthood, which means clocks trained on 30-to-80-year-olds may systematically miss the mark in a 5-year-old. Evaluations of seven different clocks across multiple tissue types from healthy children confirmed that clocks specifically trained on pediatric data performed best.¹⁵

One pediatric-specific clock, the PedBE clock, was developed using cheek cell DNA from over 1,700 children aged 0 to 20. It selected 94 methylation sites and achieved a median error of just 0.35 years in its test dataset.¹⁶ That level of precision in children opens up clinical applications, including monitoring whether pediatric diseases are accelerating biological aging in ways that chronological milestones would not reveal.

Forensic Applications

When police find biological evidence at a crime scene but have no suspect, DNA methylation can help narrow the search. The technique is part of a broader approach known as forensic DNA phenotyping, which uses genetic and epigenetic markers to infer characteristics like ancestry, appearance, and age when conventional identification has failed.¹⁷ Knowing that a suspect is probably in their twenties rather than their fifties can be a significant investigative lead.

Practical forensic use demands that the test work with small, degraded samples. Researchers have developed methods that maintain accuracy even with as little as five nanograms of starting DNA, roughly the amount you might extract from a small bloodstain or a few skin cells on a surface.⁴ Multiplex strategies that test several methylation sites simultaneously from tiny samples have further broadened what is possible in casework.¹⁸

Population Differences and Ancestry

Most widely used epigenetic clocks were built predominantly from samples of European descent, which raises a fair question: do they work as well for everyone? Population-specific clocks are now being developed. One designed specifically for East Asian populations, using methylation data from that group, achieved an average error of about six to seven years in testing.¹⁹ That is somewhat less precise than the best general-purpose clocks, partly because population-specific datasets tend to be smaller. The broader principle holds: ancestry influences the methylation landscape enough that a one-size-fits-all clock may introduce small but systematic biases, and population-tailored models can help correct for that.

Can Epigenetic Age Be Reversed?

If lifestyle factors can accelerate your DNA age, the natural follow-up is whether you can push it back. There is preliminary evidence that yes, at least partially, you can. A small pilot trial tested an eight-week diet and lifestyle program and found that participants scored an average of about 3.2 years younger on the Horvath clock compared to a control group.²⁰ The effect was modest and the trial was small, so it should be treated as suggestive rather than definitive.

A separate trial using a pharmacological protocol intended to regenerate the thymus observed a mean decrease of about 2.5 years in epigenetic age over the course of a year, with the effect accelerating toward the end of the treatment period. The decrease persisted for at least six months after participants stopped the treatment.²¹ Laboratory work on partial cellular reprogramming has shown that epigenetic age can be reset without erasing a cell’s identity, which is a crucial distinction because full reprogramming turns cells back into stem cells, an approach that could be dangerous in a living organism.²² Whether any of this translates into actually living longer remains unproven.

Why Methylation Changes Happen in the First Place

A persistent mystery has been what drives the methylation shifts that clocks rely on. One emerging explanation involves somatic mutations, the random genetic errors that accumulate in our cells as we age. A large study analyzing blood samples from over 200,000 individuals found that the frequency and size of these mutations increase steadily with age.²³ Strikingly, the genomic locations where mutations pile up with age tend to be the same locations where methylation patterns are most predictive of age. This suggests that the ticking of epigenetic clocks is not an independent program but is closely coupled to the gradual accumulation of random DNA damage.²⁴

Mitochondrial DNA provides another lens on the same process. The number of mitochondrial DNA copies in a cell decreases with age, dropping by roughly 0.4 copies per year on average, while the number of mutations in the mitochondrial genome increases. People over 70 carry about 60% more mitochondrial mutations than people under 40.²⁵ These mitochondrial changes are tissue-specific: skeletal muscle shows a pronounced decline in copy number with age, especially in men, while liver shows the opposite trend.²⁶ Mitochondrial markers have not reached the precision of methylation clocks for age estimation, but they offer independent confirmation that aging leaves deep, measurable marks throughout the genome.

Beyond Methylation and Into Gene Activity

Methylation is the most developed molecular clock, but it is not the only one. The activity levels of genes, measured by how much RNA they produce, also shift with age. Researchers have built transcriptomic age predictors that estimate age from gene expression patterns. One such model, trained on over 130 skin cell samples spanning ages 1 to 94, achieved a median error of four years.²⁷ A broader tool called RNAAgeCalc was developed using multi-tissue data to create both cross-tissue and tissue-specific age predictors from gene expression.²⁸

Transcriptomic clocks capture different biological information than methylation clocks. Gene expression reflects what a cell is actively doing at a given moment, including inflammatory signaling, stress responses, and metabolic activity, while methylation captures a more stable, cumulative record. The two approaches are complementary. A person whose methylation clock reads normal but whose gene expression clock reads old might be in the early stages of a disease process that has not yet left a permanent epigenetic mark.

Aging Clocks for Wild Animals

The same methylation-based principles work across species, which has opened up practical applications in wildlife biology. A universal mammalian clock, built from nearly 12,000 methylation profiles spanning 185 species and 59 tissue types, estimates age with a correlation above 0.96.²⁹ For species that are difficult or impossible to age by conventional means, this is transformative.

Toothed whales and dolphins are a good example. Aging wild cetaceans historically required counting growth layers in teeth, which demands either a dead animal or an invasive biopsy. A methylation clock built from skin and blood samples of nine species of toothed whales achieved a correlation of 0.94 with known age and was then applied to estimate ages of wild Norwegian killer whales from skin dart biopsies.³⁰ Similar approaches have been developed for mammals, birds, and fish, making epigenetic age estimation an increasingly standard tool in conservation biology.³¹

Technical Limits Worth Knowing About

For all the progress, the technology is not yet plug-and-play. The measurement platforms themselves introduce noise. An analysis of technical replicates, the same sample measured twice, found that fewer than a quarter of the methylation sites on a widely used array had acceptable reproducibility with raw data. Preprocessing steps can improve that, but the figure still topped out around 39%.³² Clock models work despite this because they aggregate hundreds of sites, and the noise at any single site tends to wash out in the average. But the underlying imprecision at the individual-site level means that predictions for any single person always carry real uncertainty, even when the average error across a population looks impressively low.

There is also a fundamental tension in what these clocks measure. A clock designed to estimate chronological age as precisely as possible will, by definition, be less sensitive to the biological deviations that make methylation age clinically interesting. A clock designed to capture biological aging will be less precise at guessing birthdays. No single clock does both jobs perfectly, which is why the field now has multiple generations of clocks optimized for different purposes.