Epigenetic Clocks Explained: How to Measure Your Biological Age

What Is an Epigenetic Clock?

When I first encountered the concept of an epigenetic clock back in 2013, I remember thinking it sounded like science fiction. The idea that a relatively simple chemical mark on your DNA — not the DNA sequence itself, but a tiny methyl group attached to it — could predict your risk of disease, your remaining lifespan, and even how your body is ageing at a cellular level felt almost impossibly ambitious. More than a decade later, I’ve read hundreds of papers in this field, had my own biological age tested multiple times, and I can tell you: this is one of the most genuinely exciting developments in the entire science of longevity.

An epigenetic clock is a computational tool — an algorithm, really — that uses patterns of DNA methylation at specific sites across your genome to estimate your biological age. It doesn’t tell you how many years you’ve been alive (that’s your chronological age). It tells you how old your cells and tissues appear to be based on molecular evidence. The gap between the two numbers is, I’d argue, the most actionable piece of data most of us can get about our long-term health trajectory.

To understand this properly, you need to understand a little about the hallmarks of ageing and where epigenetics fits within that framework. Epigenetic alteration is itself one of those hallmarks — a recognised driver of the ageing process, not merely a passive marker of it. That distinction matters enormously for how we interpret what these clocks are measuring.

How DNA Methylation Works

Your genome contains roughly 28 million sites where a cytosine base (the “C” in the ACGT alphabet of DNA) sits next to a guanine base (“G”) — these are called CpG sites. At many of these locations, an enzyme called a DNA methyltransferase (DNMT) can attach a methyl group (–CH₃) to the cytosine. When this happens, we say the site is “methylated.” When the methyl group is absent, it’s “unmethylated.”

Methylation generally silences nearby genes. Unmethylated promoter regions tend to allow gene expression. This on/off switching is how your liver cells and your skin cells, despite containing identical DNA sequences, express completely different sets of proteins and behave so differently from one another.

Here’s where ageing enters the picture. As we age, methylation patterns across the genome change in remarkably consistent, predictable ways. Some sites that were methylated in youth become unmethylated over time; others gain methylation they shouldn’t have. This isn’t random drift — it follows a programme-like trajectory that is conserved across individuals and even across species. That consistency is precisely what makes it measurable, and measurable means potentially modifiable.

The Horvath Clock Explained

In 2013, biostatistician Steve Horvath at UCLA published what became the foundational paper in this field. He trained a machine-learning model on methylation data from 8,000 samples across 51 tissue types and identified 353 CpG sites whose combined methylation levels predicted chronological age with striking accuracy — a correlation of about 0.96 across diverse tissues. The resulting tool became known as the Horvath clock, and it changed the field overnight.

What made Horvath’s finding remarkable wasn’t just the accuracy. It was the universality. The same 353 sites predicted age whether you were measuring blood, brain tissue, breast tissue, or saliva. This suggested these methylation changes aren’t simply a consequence of tissue-specific function — they reflect something more fundamental about the ageing process itself. Horvath himself has described it as an “epigenetic pacemaker.”

The original Horvath clock paper (PMID: 24138928) has been cited over 10,000 times and remains the most influential single study in the biological age measurement field. I’ve read it multiple times, and I’m still struck by the clarity of the finding: your body keeps a molecular clock, and we now know how to read it.

Generations of Epigenetic Clocks

The Horvath clock was just the beginning. Researchers quickly recognised that a clock trained purely to predict chronological age has limitations — after all, if it perfectly predicted chronological age, it would be telling us nothing we didn’t already know from a birth certificate. What we really want is a clock that predicts health outcomes and mortality risk. This realisation drove the development of successive generations of clocks.

The second major clock, published by Gregory Hannum and colleagues in 2013 (PMID: 23177740), used blood-specific methylation data and 71 CpG sites. It performed similarly to Horvath’s tool in blood but not across tissues.

The more clinically relevant breakthrough came with second-generation clocks trained directly on health and mortality data rather than chronological age. The PhenoAge clock (Levine et al., 2018; PMID: 29676998) incorporated nine clinical biomarkers — including albumin, creatinine, glucose, and white blood cell count — alongside age, then identified CpG sites associated with this composite “phenotypic age.” PhenoAge acceleration (being biologically older than your chronological age predicts) turns out to be a stronger predictor of mortality, cancer risk, and cardiovascular disease than either the Horvath or Hannum clocks.

Shortly after, the GrimAge clock (Lu et al., 2019; PMID: 30669119) was developed using plasma proteins associated with mortality. GrimAge is currently considered the gold standard for mortality prediction from methylation data — biological age acceleration measured by GrimAge has been shown to predict time-to-death, coronary heart disease, and cancer independently of conventional risk factors.

Most recently, third-generation clocks like DunedinPACE (Belsky et al., 2022; PMID: 35029144) move away from estimating a static biological age and instead measure the pace of ageing — essentially, how fast you are ageing right now. DunedinPACE was trained on longitudinal data from the Dunedin cohort in New Zealand, tracking individuals over decades, and it measures decline in 19 organ systems simultaneously. A DunedinPACE score of 1.0 means you’re ageing at the average rate; below 1.0 means slower; above 1.0 means faster.

What the Research Actually Shows

I want to be direct with you here, because there’s a tendency in longevity media to overstate what epigenetic clocks can and cannot do. Let me separate the genuinely robust findings from the areas where I think we need more caution.

What’s well-established: Epigenetic age acceleration — being biologically older than your chronological age according to these algorithms — is consistently associated with increased all-cause mortality, cardiovascular disease risk, cancer incidence, and cognitive decline. The GrimAge clock in particular has been validated in multiple independent cohorts. A meta-analysis by Kresovich and colleagues (PMID: 33601204) confirmed that epigenetic age acceleration across multiple clocks is associated with cancer risk. These are not trivial associations — they hold after adjusting for smoking, BMI, alcohol, and socioeconomic status.

What’s more uncertain: Whether interventions that reduce your epigenetic age score actually extend your healthy lifespan in humans. We have compelling data showing that things like diet, exercise, and stress reduction correlate with lower epigenetic age. We have some short-term intervention studies — most notably Fahy and colleagues’ TRIIM trial (PMID: 31931803), which showed a reversal of approximately 2.5 years of epigenetic age in a small cohort of nine men using a combination of growth hormone, metformin, and DHEA. But “reversal of a clock score” is not the same as “extension of healthy human lifespan.” The clocks are proxies. Very good proxies, but proxies nonetheless.

Variability between tests: If you take your saliva sample on Monday and again on Friday, your score will not be identical. There is biological noise in these measurements, and different clocks can give different “biological ages” from the same sample. I’ve had test results that differed by four years between two reputable services, using samples taken the same week. This doesn’t mean the technology is worthless — it means you should treat a single result as a data point, not a verdict.

Biological Age vs Chronological Age

Your chronological age is simply the number of years since you were born. It’s fixed, unavoidable, and utterly useless as a health metric on its own. Two 55-year-olds can have radically different cardiovascular function, cognitive sharpness, inflammatory load, and cellular repair capacity. We all know this intuitively — we’ve all met the 70-year-old who seems decades younger, and the 45-year-old who seems to have aged prematurely.

Biological age attempts to capture the actual state of your physiology. Epigenetic biological age is one way to estimate this, but it’s worth knowing it isn’t the only way. Telomere length, inflammatory biomarker panels, proteomics-based clocks (like the recent SomaScan-based aging measures), and functional assessments like grip strength, VO2 max, and gait speed all provide complementary information. Epigenetic clocks are arguably the most molecular and mechanistic of these measures, operating closer to the actual causal machinery of ageing.

The relationship between chronological and biological age is not fixed across your lifetime. Evidence suggests that biological ageing is not linear — it appears to accelerate in two waves, around ages 34 and 60, based on proteomics data from the Stanford Human Ageing Project. Epigenetic clocks tend to show more gradual trajectories, but the principle is the same: the gap between your biological and chronological age can widen or narrow depending on your exposome — the totality of environmental exposures across your life.

How to Test Your Biological Age

Consumer-facing epigenetic age tests have become widely available in the past few years. Most use a saliva or dried blood spot sample that you collect at home and post to a laboratory. The methylation array used by the vast majority of commercial and research labs is the Illumina EPIC array (or its predecessor, the 450K array), which measures methylation at approximately 850,000 CpG sites simultaneously. From that data, the relevant clock algorithms are applied.

Services I’ve personally used or reviewed include TruDiagnostic (which runs multiple clocks including DunedinPACE and GrimAge), Elysium Index (which uses PhenoAge and Horvath), and myDNAge (which uses a version of Horvath’s multi-tissue clock). Prices typically range from £150 to £400 per test, with lower rates if you subscribe to track changes over time.

My practical advice: if you’re going to test, commit to testing at least twice, six to twelve months apart, and standardise your conditions. Take your sample at the same time of day, after a similar dietary period, when you’re not acutely ill or under unusual stress. Single snapshots are interesting; longitudinal tracking is genuinely informative.

What Accelerates Epigenetic Ageing

The research literature is fairly consistent on the major drivers of epigenetic age acceleration. These aren’t surprising — they broadly overlap with general health risk factors — but the epigenetic data gives us a molecular window into why these exposures are damaging at a cellular level.

• Smoking: The strongest and most replicated environmental driver of epigenetic age acceleration. Smoking-associated methylation changes are detectable at hundreds of CpG sites and persist for years after cessation, though partial reversal does occur.
• Excess adiposity: BMI and visceral fat are consistently associated with faster epigenetic ageing, particularly on clocks that incorporate metabolic biomarkers like PhenoAge.
• Chronic psychological stress: Adverse childhood experiences (ACEs) and ongoing psychological stress are associated with accelerated methylation ageing, partly mediated through glucocorticoid signalling and inflammation.
• Sleep disruption: Both short sleep duration and poor sleep quality are associated with higher epigenetic age, consistent with the well-established role of sleep in cellular repair and circadian regulation of methylation.
• Ultra-processed food consumption: Recent data links high UPF intake to accelerated epigenetic ageing, likely through combined effects of inflammation, oxidative stress, and gut microbiome disruption.
• Sedentary behaviour: Physical inactivity accelerates epigenetic ageing independently of BMI. Moderate-to-vigorous exercise, conversely, is one of the most consistently protective factors identified.
• Alcohol: Heavy alcohol consumption is associated with accelerated GrimAge in particular, reflecting its widespread impact on hepatic, cardiovascular, and immune function.

What Can Slow or Reverse Epigenetic Ageing

This is the area that generates the most excitement — and the most hype. I want to be honest about the quality of evidence behind each category, because there’s a significant difference between “associated with lower biological age in cross-sectional studies” and “causally reduces epigenetic ageing in randomised controlled trials.”

Exercise (strong observational evidence, some RCT data): Multiple studies show lower epigenetic ages in physically active individuals. A study by Spartano and colleagues using the Framingham Heart Study cohort linked higher physical activity to lower GrimAge acceleration. Short-term exercise intervention trials have shown modest but measurable reductions in epigenetic age acceleration.

Caloric restriction and fasting (compelling but limited human data): Animal data is robust — caloric restriction consistently slows epigenetic ageing in rodents and primates. Human data is more limited. The CALERIE trial showed modest epigenetic age benefits from 25% caloric restriction over two years. Time-restricted eating is promising but human trial data specifically on methylation clocks remains sparse.

Mediterranean diet and dietary quality (moderate evidence): Higher diet quality scores correlate with lower epigenetic age across multiple studies. A randomised trial (PMID: 31931803 and related work from the TRIIM study) demonstrated clock reversal, but the multi-component nature of interventions makes isolating dietary effects difficult.

Specific supplements and pharmacological agents (early/speculative): Metformin, rapamycin, and NAD+ precursors (NMN, NR) are all being studied for effects on epigenetic ageing. Some researchers following the David Sinclair protocol are specifically tracking clock scores to assess efficacy. The evidence for meaningful clock reversal from any single supplement in humans remains preliminary. I take NMN personally and track my DunedinPACE scores, but I won’t claim a causal link without better trial data.

Psychological interventions and social connection (emerging): Loneliness and social isolation are associated with epigenetic age acceleration. Conversely, mindfulness-based interventions have shown small but statistically significant reductions in epigenetic age in some trials. This is an underappreciated area.

Epigenetic Clock Comparison Table

Different clocks measure different things and have different strengths. Here’s my summary of the major validated clocks you’re likely to encounter when reading research or ordering a commercial test.

Common Myths and Misconceptions

After years of following this field, I’ve noticed the same misconceptions recurring — both in popular media and in online longevity communities. Let me address the ones that bother me most.

Myth 1: “My epigenetic age IS my biological age.” Epigenetic clocks are one measure of biological age, not the definitive measure. Biological age is a multidimensional concept. A person with a young epigenetic age can still have significant arterial stiffness, poor lung function, or declining muscle mass. Use clock scores as one input alongside other assessments.

Myth 2: “Reducing my clock score will definitely make me live longer.” We don’t yet know this with certainty. Clock scores predict health outcomes at the population level. Whether optimising your personal score via interventions translates to extended lifespan in the way we hope is an empirical question that has not been definitively answered. I’m cautiously optimistic — but intellectually honest about the gap in the evidence.

Myth 3: “Epigenetic changes from lifestyle are permanent.” Many epigenetic changes are reversible. Smoking-associated methylation changes partially reverse after cessation. Exercise induces rapid epigenetic changes at thousands of CpG sites. This reversibility is part of what makes the field so exciting from a therapeutic standpoint.

Myth 4: “A younger result means I’m healthy across the board.” Clocks trained on blood measure blood-cell epigenetics. They may not capture tissue-specific ageing in organs like the brain or prostate. Some researchers are working on tissue-specific clocks precisely because systemic clocks can miss localised accelerated ageing.

Myth 5: “The Horvath clock is the best one to use.” The Horvath clock was transformative for its time, but for practical health prediction, GrimAge and DunedinPACE now outperform it. The choice of clock should depend on what question you’re asking. For longevity prediction: GrimAge. For current pace of ageing: DunedinPACE. For developmental biology research: Horvath.

Action Steps: What to Actually Do

I don’t want to leave you with just theory. If you’re a health-conscious adult who wants to engage meaningfully with this technology, here is the practical framework I’d recommend — the same one I follow myself.

1. Get a baseline test. Choose a service that runs at minimum GrimAge and DunedinPACE (TruDiagnostic is currently my preferred option for this). Request your raw methylation data export if possible. This is your starting reference point.
2. Don’t optimise for the clock — optimise for health. The interventions most consistently associated with lower epigenetic age are the same ones associated with better health across every metric: regular vigorous exercise, high diet quality (especially vegetables, legumes, oily fish, minimal UPFs), 7–9 hours of sleep, not smoking, limited alcohol, and managing chronic stress. Do these first, before considering anything more exotic.
3. Retest in 6–12 months. A single test is a snapshot. Longitudinal tracking under consistent conditions gives you signal above the noise. If your DunedinPACE has improved after six months of consistent lifestyle changes, that is meaningful data.
4. Interpret in context. If your clock score is older than your chronological age, don’t panic — but do treat it as a prompt to examine the likely drivers: sleep quality, weight, exercise frequency, stress levels, smoking. The modifiable factors are well-characterised.
5. Stay current with the research. This field is moving rapidly. New clocks, new interventions, and new validation data are published regularly. The clocks I’ve described here represent the state of the art in 2024, but the landscape in 2026 may look quite different.

Frequently Asked Questions