Epigenetic Clocks Explained: How Methylation Predicts Biological Age

In 2013, a biostatistician at UCLA named Steve Horvath published a paper that changed how researchers think about aging. By analyzing DNA methylation patterns at 353 specific sites in the genome, he could predict a person's chronological age with a median absolute error of 3.6 years, across 51 different tissue types and cell types. The tool became known as the Horvath clock, and it launched an entire field.

A decade later, we have at least five major epigenetic clocks, each designed to answer a different question. Most people tracking their biology today have access to these tests but not a clear framework for what each clock actually measures. That's what this page is for.

---

DNA Methylation 101: CpG Sites and What Changes With Age

DNA methylation is the addition of a methyl group (CH3) to the cytosine base of DNA, almost always at locations called CpG sites, positions in the genome where a cytosine nucleotide is followed immediately by a guanine. There are roughly 28 million CpG sites in the human genome, and about 1 to 2 percent of them change their methylation status predictably with age.

These systematic changes come in two flavors:

Hypermethylation means a site gains methylation with age. When this occurs at the promoter regions of tumor suppressor genes, it can silence those genes, which is one reason cancer risk rises with age.

Hypomethylation means a site loses methylation with age. When this occurs near inflammatory cytokine genes, it can increase their expression, contributing to the chronic low-grade inflammation often called "inflammaging."

About 60 percent of the Horvath clock CpG sites lose methylation with age and 40 percent gain it. This bidirectionality is important: it means a simple "methylate everything more" strategy would not work. What matters is whether specific sites are at the right level for a younger biological state. It also explains why interventions like the Fitzgerald 2021 RCT saw no net increase in overall methylation, but a meaningful repositioning of the pattern.

Methylation is measured from tissue samples, most commonly blood, saliva, or cheek swabs, using microarray platforms (the Illumina EPIC array is the current standard) that profile methylation levels at hundreds of thousands of CpG sites simultaneously. The resulting data is fed into the relevant clock algorithm to produce an age estimate.

---

The Five Major Epigenetic Clocks

1. Horvath 2013 (Multi-Tissue Clock)

The original. Trained on 8,000 samples from 82 datasets spanning 51 healthy tissues, the Horvath clock uses 353 CpG sites and was designed to estimate chronological age as accurately as possible across virtually any tissue type. It achieves this with a median absolute error of 3.6 years.

The multi-tissue design is its primary strength: consistent results whether the sample comes from blood, brain, liver, or kidney. Its primary limitation is that it was calibrated to predict chronological age rather than biological health or mortality risk. This means it captures something fundamental about cellular aging but is not the most sensitive tool for detecting lifestyle interventions or predicting who will die sooner.

2. Hannum 2013 (Blood Clock)

Published in the same year as Horvath's clock, Hannum's clock was trained on whole blood samples. It uses 71 CpG sites and was one of the first to show that blood-based methylation age predicts all-cause mortality independently of chronological age. The blood-specific design makes it more sensitive to immune system aging than the multi-tissue clock, but it cannot be applied to non-blood tissue samples.

3. PhenoAge (Levine, 2018)

Developed by Morgan Levine and colleagues at Yale, PhenoAge represents the first major shift toward what researchers call second-generation clocks. Rather than training on chronological age, PhenoAge was trained on "phenotypic age": a composite of nine clinical biomarkers (albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean corpuscular volume, red cell distribution width, alkaline phosphatase, and white blood cell count) that are themselves predictive of mortality. The result is a clock that predicts mortality and disease risk substantially better than either Horvath or Hannum.

PhenoAge acceleration (the gap between your PhenoAge and your chronological age) has been associated with frailty, cancer, cardiovascular disease, and neurodegenerative disease risk.

4. GrimAge (Lu, 2019)

GrimAge, developed by Ake Lu and Steve Horvath, is trained on DNA methylation surrogates of seven plasma proteins associated with mortality and morbidity, plus smoking pack-years. The name is intentional: it is the strongest predictor of mortality among all epigenetic clocks. In a 2025 retrospective cohort study of 1,942 NHANES participants, GrimAge and its updated version GrimAge2 showed stable, approximately linear associations with all-cause, cardiac, and cancer mortality, the only clocks in the analysis to demonstrate this pattern consistently.

GrimAge is a second-generation clock. It does not just estimate how old you look epigenetically; it estimates how much physiological wear your body has accumulated in the systems that matter most for survival. GrimAge2 (2022) updates the protein biomarkers to include CRP and HbA1c surrogates, improving its relevance to cardiometabolic aging.

5. DunedinPACE (2022)

DunedinPACE (Pace of Aging Computed from the Epigenome) is the most conceptually different of the five. Rather than giving an age estimate, it gives a rate of how many biological years are passing per calendar year. A DunedinPACE score of 1.0 means you are aging at the population-average rate. A score of 1.2 means you are aging 20 percent faster than average. A score of 0.85 means you are aging at 85 percent of the average rate.

The clock was developed from the Dunedin longitudinal cohort, tracking the same New Zealand birth cohort from birth to midlife. This design lets it measure actual aging change rather than cross-sectional differences between old and young people. DunedinPACE is the most sensitive clock for detecting intervention effects. In the DO-HEALTH trial (2025), DunedinPACE was one of the three clocks that responded to omega-3 supplementation. It is also the most responsive to short-term lifestyle changes.

When I got my DunedinPACE score back at 1.15 on my first TruDiagnostic test, that number hit differently than the biological age estimate. It meant I was accumulating 1.15 biological years for every calendar year. Not catastrophic, but not where I wanted to be. Fourteen months later, after prioritizing sleep consistency and dropping my Zone 2 sessions to actually stay in Zone 2, the score moved to 1.04. That's the kind of directional feedback that makes these tests worth the money.

---

First, Second, and Third Generation: What the Labels Mean

Researchers use a generational taxonomy to describe the evolution of these tools:

First generation (Horvath 2013, Hannum 2013): Trained to predict chronological age. Highly accurate at that task, but show weak and inconsistent associations with morbidity and mortality beyond what chronological age already predicts.

Second generation (PhenoAge, GrimAge, GrimAge2): Trained on mortality risk or clinical phenotypes rather than calendar age. Show much stronger associations with disease and death. GrimAge is the most powerful mortality predictor in this class.

Third generation (DunedinPACE): Trained on longitudinal aging change in a single cohort rather than cross-sectional age differences. Measures rate of aging rather than state. Most responsive to lifestyle intervention.

For most health applications, second- and third-generation clocks provide the most actionable information.

---

How a Sample Is Analyzed

The consumer process typically follows these steps:

1. You collect a blood spot (finger-prick) or saliva sample at home using a kit sent by the testing company.
2. DNA is extracted from the sample and applied to an Illumina EPIC methylation microarray, which measures beta values (0 to 1) at approximately 900,000 CpG sites.
3. Raw beta values are normalized using pipelines such as SeSAMe or Noob to correct for technical variation between samples.
4. The normalized values at clock-relevant CpG sites are fed into the specific clock algorithm to produce an age estimate or pace-of-aging score.
5. Results are returned through a consumer portal, usually within two to four weeks.

The critical quality control step is normalization. Different normalization methods can produce different absolute numbers from the same DNA. When tracking your own change over time, use the same testing company and the same normalization pipeline so that changes reflect biology rather than technical variation.

---

Accuracy and Limitations

Epigenetic clocks are accurate biomarkers of population-level aging but show individual-level variability. For first-generation clocks, the median absolute error is roughly 3.6 years. For second-generation clocks, the emphasis shifts from age estimation to risk prediction, and individual scores can differ substantially from lab to lab depending on normalization.

Key limitations to understand:

• Clock scores capture the average methylation state of millions of cells in a tissue sample, not any single cell. This matters because cancer cells and immune cells have dramatically different methylation profiles.
• Smoking accelerates GrimAge substantially. GrimAge was specifically trained to capture this. A former smoker's GrimAge will often look older than their Horvath age.
• Clocks trained on one population may not generalize equally to all ethnicities. Calibration studies in diverse populations are ongoing.
• No consumer epigenetic clock result has yet been validated as a clinical diagnostic. These are research-grade biomarkers, not medical tests.

---

Consumer Tests Available

TruDiagnostic (TruAge): The most comprehensive consumer option. Reports multiple clocks including PhenoAge, GrimAge2, and DunedinPACE from a finger-prick blood sample. Testing typically costs $300 to $500 per panel. TruDiagnostic was the lab used in the Fitzgerald 2021 women's case series.

Elysium Health (Index): Uses a saliva sample and reports a biological age based on proprietary methylation analysis. More consumer-accessible in terms of price and user experience, but reports fewer clock types than TruDiagnostic.

GlycanAge: Measures biological age through immunoglobulin G glycan patterns in blood rather than DNA methylation. Validated against cardiovascular outcomes and inflammation markers. A meaningfully different biomarker than methylation clocks, though both track biological aging. Priced around $200 to $300 per test.

For tracking the specific interventions discussed in our guide to lowering biological age, GrimAge2 and DunedinPACE provide the most signal. Retesting every three to six months allows enough time to accumulate detectable biological change.

---

FAQ

Are epigenetic clocks the same as genetic age tests like 23andMe? No. Genetic tests (SNP arrays) read your DNA sequence, which does not change. Epigenetic clocks read methylation patterns on top of that sequence, which do change with age, environment, and behavior. They answer completely different questions.

Can one test tell me my biological age accurately? A single test gives you a baseline, not certainty. Absolute scores vary depending on the normalization method, the specific clock used, and individual biological variation. Repeat measurements over time, tracking your own trajectory, are more informative than any single result.

Why do my different clock scores disagree? Each clock was trained to answer a different question. Horvath estimates chronological age. GrimAge predicts mortality risk. DunedinPACE measures current aging speed. A 45-year-old could have a Horvath age of 41, a GrimAge suggesting elevated mortality risk, and a DunedinPACE showing normal aging pace. These are complementary, not contradictory, perspectives.

Does smoking age the epigenetic clock? Yes, and substantially. GrimAge was specifically trained on smoking pack-years as a component. Quitting smoking is associated with partial reversal of smoking-related epigenetic age acceleration over time.

How do epigenetic clocks relate to wearable data? Wearables like Oura Ring and Garmin measure proxies for the same physiological systems that epigenetic clocks track, including HRV, sleep quality, and recovery, on a daily basis. Pairing daily wearable data with periodic clock tests through a platform like Vaya Chat at Vidaya lets you connect short-term behavioral patterns to longer-term biological trends.

What is clock "acceleration" and why does it matter? Epigenetic age acceleration is the gap between your clock score and your chronological age. A positive number means your biology is older than your calendar age. Acceleration is associated with higher risks of cancer, cardiovascular disease, neurodegeneration, and all-cause mortality. Reducing acceleration is the goal of most biological age interventions.

---

---

Epigenetic clocks give you a molecular window into how your body is aging. But a single score every few months is limited without the daily context of what your sleep, stress, and activity look like in between. Vaya Chat at Vidaya connects your wearable data and health metrics so you can build the full picture. Plans start at $10 per month.

Related reading: How to Lower Your Biological Age | Understanding Your Biological Age | Oura Ring Data Explained | Garmin Data Explained