How an Epigenetic Clock Measures DNA Methylation and Risk

How an Epigenetic Clock Works

An epigenetic clock measures chemical tags called methyl compounds attached to DNA to estimate biological aging. According to the Cleveland Clinic, DNA methylation occurs when small groups of chemicals attach to cellular DNA without altering the underlying genetic sequence. Dr. Saadia Hussain describes DNA methylation as one of the most common markers used today for biological age, an idea first published in 2013.

These chemical marks sit at cytosine-phosphate-guanine sites, commonly termed CpG sites. The clocks on this page read methylation at 71 to 1,030 selected CpG sites and combine those readings into one score.

In 2013, Horvath regressed a transformed version of chronological age on CpG sites using an elastic net regression model. That mathematical model automatically selected 353 CpGs. Horvath reported that the weighted average of these selected sites amounts to an epigenetic clock. Later models such as GrimAge and DunedinPACE also use elastic-net models. As Levine and colleagues observed in 2018, early epigenetic biomarkers used chronological age as a surrogate for biological age. Newer clocks train their algorithms on physical decline, disease incidence, or mortality outcomes instead of birth years.

Horvath, Hannum, PhenoAge, GrimAge and DunedinPACE by Generation

Epigenetic clocks fall into three developmental generations based on whether their algorithms train on calendar age, clinical outcomes, or longitudinal organ decline.

Horvath's 2013 multi-tissue clock evaluated 8,000 samples across 82 Illumina DNA methylation datasets covering 51 healthy tissues and cell types. The test-set median error was 3.6 years. Predicted DNA methylation age differed from chronological age by less than 3.6 years in 50% of subjects. DNA methylation age was close to zero for embryonic and induced pluripotent stem cells. It correlated with cell passage number and applied successfully to chimpanzee tissues. Horvath reported a highly heritable measure of age acceleration across populations. Across 6,000 cancer samples from 32 datasets, all 20 cancer types showed age acceleration averaging 36 years.

In 2013, Hannum and colleagues evaluated more than 450,000 CpG markers in whole blood from 656 individuals aged 19 to 101. Their model selected 71 methylation markers. The correlation between chronological age and predicted age reached 96% with an error of 3.9 years. Validation testing produced a 91% correlation and an error of 4.9 years. Hannum and colleagues noted that aging rates are impacted by gender and genetic variants.

Second-generation models replaced chronological age targets with clinical disease risks. In 2018, Levine and colleagues published DNAm PhenoAge using a two-step training methodology. Step one calculated phenotypic age from chronological age and nine clinical biomarkers using mortality records from 9,926 adults in NHANES III. Those blood markers included albumin, creatinine, serum glucose, C-reactive protein, lymphocyte percent, mean cell volume, red cell distribution width, alkaline phosphatase, and white blood cell count. Step two regressed that phenotypic age on blood methylation, selecting 513 CpGs. The authors reported that PhenoAge strongly outperforms previous measures for predicting all-cause mortality, cancers, healthspan, physical functioning, and Alzheimer's disease. Built from whole blood, it correlates strongly with age in every tissue and cell tested. A provisional patent application names two authors as inventors.

In 2019, Lu and colleagues constructed DNAm GrimAge using 1,030 unique CpGs. GrimAge combines seven DNA methylation surrogates of plasma proteins, including PAI-1 and growth differentiation factor 15, with an estimate of smoking pack-years. The authors reported that GrimAge stands out among existing epigenetic clocks for predicting time-to-death, coronary heart disease, and cancer. Cox regression analysis yielded a P-value of 2.0E-75 for time-to-death, 6.2E-24 for coronary heart disease, and 1.3E-12 for cancer. A provisional patent application names two authors as inventors.

Third-generation testing tracks the current rate of physiological deterioration. In 2022, Belsky and colleagues introduced DunedinPACE using data from the Dunedin Study 1972-1973 birth cohort. The researchers tracked within-individual decline across 19 indicators of organ-system integrity at four time points spanning two decades. They distilled those measurements into a single-time-point blood test selecting 173 CpG sites through elastic-net regression. Model selection was restricted to probes with acceptable test-retest reliability. DunedinPACE is scaled to one year of biological aging per calendar year, with a cohort mean of 1 and a standard deviation of 0.29. The test-retest intraclass correlation coefficient was 0.96 across 36 replicate blood samples. The authors reported effect sizes similar to GrimAge, with incremental prediction beyond GrimAge for morbidity, disability, and mortality. Several authors are named on a university invention licensed to a commercial entity.

What Epigenetic Age Acceleration Means

Epigenetic age acceleration quantifies the difference between an individual's estimated DNA methylation age and their chronological age. In 2015, Marioni and colleagues called this discrepancy between DNA methylation age and chronological age Δage. Lu and colleagues later generated AgeAccelGrim by adjusting DNAm GrimAge for chronological age. In 2016, Chen and colleagues evaluated intrinsic epigenetic age acceleration, defined as the residual from regressing epigenetic age on chronological age and blood cell counts. Their analysis also documented extrinsic epigenetic age acceleration and Hannum-based variants.

These measurements predict health outcomes across large human cohorts. Marioni and colleagues examined four cohorts of older adults. A 5-year higher Δage was associated with a 21% higher mortality risk after adjusting for age and sex. That association remained at 16% after further statistical adjustments. Chen and colleagues evaluated 13 cohorts comprising 13,089 individuals. All considered measures of epigenetic age acceleration were predictive of mortality, independent of chronological age. These risk statistics describe group-level associations instead of personal clinical prognoses.

A 2021 systematic review by Oblak and colleagues evaluated 156 publications and over 1,300 study findings. The authors found that the Horvath, Hannum, Levine, and GrimAge clocks tend to agree in direction of effects, but vary in size. Body mass index, HIV infection, and male sex were associated with acceleration across one or more clocks. Epigenetic age acceleration was related to mortality, cardiovascular disease, cancer, and diabetes.

How Accurate Are Epigenetic Clocks?

Epigenetic clocks predict mortality across large cohorts, but one person's repeat tests can give different ages. In 2022, Higgins-Chen and colleagues reported that technical noise produces deviations up to 9 years between replicates for six prominent epigenetic clocks, limiting their utility. The standard deviation of epigenetic age acceleration ranges from 3 to 5 years. The authors noted that indicating a person has aged 9 years due solely to technical variation would be misleading. The authors wrote that, with conventional clocks, a change measured after a treatment or lifestyle change "would be suspect." Principal-component versions of six clocks showed agreement between most replicates within 1.5 years. That paper discloses a company that licenses two authors' metrics.

Agreement between different biological age models remains low. In 2018, Belsky and colleagues evaluated 964 Dunedin Study members across 11 biological aging measures. Epigenetic clocks correlated with one another, but correlations between clocks and biomarker algorithms were low. The authors concluded that the approaches may not measure the same aspects of the aging process.

Horvath's and Hannum's clocks were trained on calendar age, while PhenoAge, GrimAge, and DunedinPACE were trained on health, mortality, or organ-decline data. In the CALERIE trial, calorie restriction slowed DunedinPACE and produced no significant change in PhenoAge or GrimAge in the same 220 adults. Cleveland Clinic notes that biological age tests are not part of routine medical care yet. Dr. Hussain emphasizes that it is not clear how reliable these tests are.

How to Get Your Epigenetic Age Tested

Consumers can access epigenetic age testing through commercial laboratory kits that analyze DNA extracted from blood or saliva samples. Horvath's multi-tissue clock was developed using data from 51 healthy tissues and cell types. PhenoAge and DunedinPACE were developed from whole blood samples. Cleveland Clinic states that direct-to-consumer online tests analyze either a blood or saliva sample.

The biological age test page compares test types, sample methods and prices. The biological age blood test guide details the nine laboratory markers in PhenoAge. The CBC blood test, comprehensive metabolic panel and CRP levels pages cover the individual PhenoAge markers.

Can Epigenetic Age Be Reversed?

Two trial analyses on this page, CALERIE and DO-HEALTH, report small effects on clock readings from calorie restriction and omega-3. In 2023, Waziry and colleagues analyzed the CALERIE trial across 220 adults without obesity. Participants were randomized to 25% calorie restriction or an ad libitum diet for two years. Calorie restriction slowed the pace of aging as measured by DunedinPACE. However, restriction did not lead to significant changes in biological age estimates from PhenoAge or GrimAge. The authors reported that treatment effect sizes were small, showing a 2% to 3% reduction in DunedinPACE. The authors say a conclusive test "will require trials with long-term follow-up" on disease and mortality.

In 2025, Bischoff-Ferrari and colleagues published findings from the DO-HEALTH trial. The post hoc analysis followed 777 older adults over three years. Omega-3 supplementation at 1 gram per day slowed PhenoAge, GrimAge2, and DunedinPACE. Combining 2,000 IU of vitamin D, omega-3, and home exercise showed additive benefits on PhenoAge only. Standardized effects ranged from 0.16 to 0.32 units, representing 2.9 to 3.8 months. The authors characterized this result as a small protective effect. One author founded a nonprofit that licenses epigenetic clock patents, including GrimAge.

In 2014, Horvath and colleagues found that obesity-linked epigenetic age acceleration in liver tissue is not reversible in the short term after rapid weight loss induced by bariatric surgery. The guide on lowering biological age and the overview of longevity supplements examine lifestyle protocols tested in published literature.

What Would Change This Answer

Long-term randomized trials showing that epigenetic clock reductions correspond with lower incidence of clinical disease would alter this assessment. Waziry and colleagues explicitly stated that establishing the clinical value of clock changes requires trials with extended follow-up for chronic illness and mortality.

Clocks that give the same result on repeat samples would also change this answer. Higgins-Chen and colleagues documented test-retest variations of up to 9 years on identical blood samples. They wrote that, with conventional clocks, a change measured after a treatment or lifestyle change "would be suspect." Principal-component algorithms reduced replicate differences to within 1.5 years in research settings. Cleveland Clinic maintains that biological age tests are not part of routine medical care yet. If you are considering an epigenetic clock test, compare the nine PhenoAge markers on the biological age blood test page with the kits on the biological age test page.