Skip to content
ContentLora

    Tip: press / anywhere to search.

    Explainer

    How epigenetic clocks and aging biomarkers work

    Epigenetic clocks estimate biological age from chemical methyl marks at hundreds of sites on DNA. Newer versions predict disease and death, or measure the pace of aging.[1][2][3] As of 2026 they are widely used in research but not validated as stand-ins for health outcomes in trials, and some clocks are noisy.[4][5]

    Editor reviewedStrict sourcingUpdated Aging and longevity biologyLife sciencesHealth and medicine

    What an epigenetic clock is

    Your DNA sequence stays almost the same throughout life, but the chemical tags attached to it change. One kind of tag, a methyl group, is added to or removed from specific spots on DNA as we age. An epigenetic clock reads these tags at many spots and converts the pattern into an age estimate.

    The first widely used clock, published in 2013, combines tags at 353 sites.[1] It was built from about 8,000 samples covering 51 tissue and cell types, so it works across most of the body.[6] One striking result: stem cells, including adult cells reprogrammed into stem cells, read as nearly zero years old.[7]

    First-generation clocks regress chronological age on CpG methylation. Horvath’s 2013 multi-tissue predictor used 353 CpGs selected from 82 Illumina array datasets covering 51 tissues and cell types.[1][6] Its estimate is near zero for embryonic and induced pluripotent stem cells, which is why clocks became a readout for partial-reprogramming experiments.[7][8]

    From age to health: newer clocks

    Predicting someone’s birthday is not very useful. Later clocks were trained to predict health instead. GrimAge, from 2019, predicted time to death better than earlier clocks.[2] DunedinPACE, from 2022, tries to measure how fast a person is aging right now. It was trained on a New Zealand group whose health was tracked over 20 years.[3]

    Second-generation clocks train on phenotypes rather than age. GrimAge combines methylation surrogates for seven plasma proteins and smoking pack-years, and outperformed earlier clocks for time-to-death.[9][2] DunedinPACE was trained on within-person decline in 19 organ-system indicators across four waves of the Dunedin cohort, restricted to reliable probes. It is associated with morbidity, disability and mortality.[3][10] In 2026 the field is moving toward multimodal and organ-level clocks; one Cell study of 2,019 adults found plasma protein clocks to be efficient proxies for physiological capacity.[11]

    How reliable are they?

    Clocks can be noisy. Measuring the same sample twice gave age estimates up to 9 years apart for six well-known clocks. A statistical fix shrank most differences to within 1.5 years.[5][12] Different clocks can also disagree about whether a treatment worked. In a two-year trial of eating 25% fewer calories, DunedinPACE showed a small slowing while GrimAge and PhenoAge did not change significantly.[13][14]

    Technical variance is a real limitation for longitudinal and trial use. Higgins-Chen and colleagues reported replicate deviations up to 9 years, reduced to under 1.5 years for most replicates with principal-component retraining.[5][12] Clock discordance in CALERIE illustrates the interpretation problem: a post hoc analysis found a small DunedinPACE effect and no significant GrimAge or PhenoAge change.[14]

    Why validation is the bottleneck

    Trials of anti-aging drugs would be far shorter if a biomarker could stand in for disease and death. The Biomarkers of Aging Consortium said in 2023 that the lack of agreed standards hinders this. A 2024 follow-up found no consensus on how to validate such biomarkers before clinical use.[15][4] A 2026 Nature Medicine review lists the potential uses: flagging high-risk people, supporting prevention, and testing whether interventions change aging.[16]

    Commercial interests overlap with the science. DunedinPACE, for example, is licensed to a testing company, and readers should weigh consumer test claims with that in mind.[17] Next in the course: how aging interventions are tested.

    Questions readers ask

    What does an epigenetic clock measure?

    It reads methylation levels at selected sites on DNA. The original Horvath clock combines 353 such sites into an age estimate.[1]

    Can a clock tell how fast someone is aging?

    DunedinPACE was built for that purpose. It is a blood test trained on two decades of measured decline in 19 indicators of organ-system health.[3]

    How accurate are these tests?

    Technical noise can make some clocks disagree by up to 9 years on the same sample. Principal-component versions brought most repeat measurements within 1.5 years.[5][12]

    Can a clock prove an anti-aging treatment works?

    Not yet. Biomarkers of aging could serve as surrogate endpoints, but there is no consensus on how to validate them for that use.[4]

    Sources

    Each numbered claim is a statement we checked against the sources listed with it. Status shows how well established it is.

    1. [1]

      The 2013 Horvath clock combines DNA methylation levels at 353 CpG sites. confirmedas of 2026-10-10

    2. [2]

      In validation data from thousands of people, GrimAge predicted time to death better than earlier epigenetic clocks. confirmedas of 2019-01-21

    3. [3]

      DunedinPACE condenses two decades of measured decline in 19 indicators of organ-system integrity in a New Zealand birth cohort into a single DNA methylation blood test of the pace of aging. confirmedas of 2022-01-14

    4. [4]

      Aging biomarkers could serve as surrogate endpoints for trials of longevity interventions, but as of 2024 there was no consensus on how they should be validated before clinical use. confirmedas of 2024-02-14

    5. [5]

      Technical noise in DNA methylation data can make six prominent epigenetic clocks disagree by up to 9 years between replicate measurements of the same sample. confirmedas of 2022-07-15

    6. [6]

      In 2013 a multi-tissue DNA methylation age predictor was built from about 8,000 samples in 82 datasets covering 51 healthy tissues and cell types. confirmedas of 2026-10-10

    7. [7]

      DNA methylation age measured by the Horvath clock is close to zero for embryonic stem cells and induced pluripotent stem cells. confirmedas of 2026-10-10

    8. [8]

      In 2020, expressing three factors (Oct4, Sox2 and Klf4, or OSK) in mouse retinal ganglion cells restored youthful DNA methylation patterns, promoted nerve regeneration and reversed vision loss in a glaucoma model and in aged mice. confirmedas of 2020-12-02

    9. [9]

      GrimAge, published in 2019, is a composite DNA methylation biomarker built from seven methylation-based estimates of plasma proteins plus a methylation-based estimate of smoking pack-years. confirmedas of 2019-01-21

    10. [10]

      DunedinPACE showed high test-retest reliability and was associated with morbidity, disability and mortality. confirmedas of 2022-01-14

    11. [11]

      A 2026 Cell study built clinical, multimodal and organ-level aging clocks from a cohort of 2,019 Chinese adults aged 18 to 91 and reported that plasma protein clocks serve as efficient proxies for systemic physiological capacity. confirmedas of 2026-05-08

    12. [12]

      Retrained principal-component versions of six clocks brought most replicate measurements within 1.5 years of each other. confirmedas of 2022-07-15

    13. [13]

      CALERIE was a randomized controlled trial in which 220 adults without obesity were assigned to 25% caloric restriction or their usual diet for two years. confirmedas of 2023-02-09

    14. [14]

      In a post hoc analysis of CALERIE, caloric restriction slowed DunedinPACE but did not significantly change biological age estimates from other clocks including PhenoAge and GrimAge, and effect sizes were small. confirmedas of 2023-02-09

    15. [15]

      A 2023 consensus paper by the Biomarkers of Aging Consortium said the lack of standards and consensus on what makes a reliable aging biomarker hinders their development and clinical validation. confirmedas of 2023-08-31

    16. [16]

      A July 2026 Nature Medicine review lists potential uses of biological aging clocks as identifying people at high risk of disease, supporting prevention or early detection, and testing whether lifestyle or interventions change the aging process. confirmedas of 2026-07-09

    17. [17]

      DunedinPACE is a Duke University and University of Otago invention licensed to the testing company TruDiagnostic, according to conflict-of-interest disclosures in a 2023 consensus paper. confirmedas of 2023-08-31

    Revision history (1)
    1. Page created.

    Created Oct 10, 2026. Last reviewed by an editor on Oct 10, 2026. Next scheduled review: Jan 10, 2027.

    Cite this page

    "How epigenetic clocks and aging biomarkers work." ContentLora, updated Oct 10, 2026. https://contentlora.com/explain/how-epigenetic-clocks-work

    Spotted an error? Suggest a correction or emailcorrections@contentlora.com.

    Keep exploring

    ☰ Contents