Loading
What changed between revision 2 and revision 3, compared over the article's markdown source with its frontmatter — so a change to status, evidence or reversibility shows up as the edit it is.
Older · revision 2
Import
Re-import of content/epigenetic-clock.md — the file changed after the initial import. Not an edit.
14.8 kB
Revision 2
Revision 3
lines 8–14 → 8–147 unchanged lines not shown
categories: ["longevity"]categories: ["longevity"]tags: ["epigenetics", "biomarkers", "aging", "dna methylation", "measurement"]tags: ["epigenetics", "biomarkers", "aging", "dna methylation", "measurement"]summary: "Statistical predictors that estimate age or mortality risk from DNA methylation patterns at selected sites in the genome."summary: "Statistical predictors that estimate age or mortality risk from DNA methylation patterns at selected sites in the genome."updated: "2026-10-28"updated: "2026-10-29"humanEvidence: "Built and validated in large human cohorts, where age acceleration predicts mortality and disease modestly; no trial has shown that moving a clock reading changes an outcome."humanEvidence: "Built and validated in large human cohorts, where age acceleration predicts mortality and disease modestly; no trial has shown that moving a clock reading changes an outcome."access: "Sold direct to consumer as mail-in methylation tests and available as a research assay on standard arrays; no clock is an approved clinical test or an accepted regulatory endpoint."access: "Sold direct to consumer as mail-in methylation tests and available as a research assay on standard arrays; no clock is an approved clinical test or an accepted regulatory endpoint."reversibility: "reversible"reversibility: "reversible"lines 29–43 → 29–4314 unchanged lines not shown
}}`````` **Epigenetic clocks** are statistical models that estimate a person's age, or their risk of death, from the pattern of methyl groups attached to cytosines at selected positions in their genome. They are the most widely used molecular measure in aging research, cheap enough to run on stored blood or saliva and accurate enough at predicting chronological age to be startling. What they measure biologically, and whether moving a clock reading means anything, remain open.**Epigenetic clocks** are statistical models that estimate a person's age, or their risk of death, from the pattern of methyl groups attached to cytosines at selected positions in their genome. They are the most widely used molecular measure in aging research: cheap enough to run on stored blood or saliva, and startlingly accurate at predicting chronological age. What they measure biologically, and whether moving a clock reading means anything, are open questions. ## What a clock measures## What a clock measures DNA methylation at CpG dinucleotides regulates gene expression and differs systematically between tissues and between young and old individuals. A clock is built by taking methylation values at hundreds of thousands of sites, measured on a commercial array, and fitting a penalised regression that predicts a target variable from a small subset of them. The selected sites are not a mechanism. They are whichever coordinates carried the most predictive signal in the training data.DNA methylation at CpG dinucleotides regulates gene expression and differs systematically between tissues and between young and old individuals. A clock is built by taking methylation values at hundreds of thousands of sites, measured on a commercial array, and fitting a penalised regression that predicts a target variable from a small subset of them. The selected sites do not amount to a mechanism; they are whichever coordinates carried the most predictive signal in the training data. For a clock trained on chronological age, the interesting quantity is not the prediction but the residual: the difference between predicted and actual age, usually called epigenetic age acceleration. A person whose blood reads five years older than their birth certificate is, on average across large cohorts, at somewhat elevated risk of death and of several chronic diseases. The effect is real and modest.For a clock trained on chronological age, the informative quantity is the residual, the difference between predicted and actual age, usually called epigenetic age acceleration. A person whose blood reads five years older than their birth certificate is, on average across large cohorts, at somewhat elevated risk of death and of several chronic diseases. The effect is real but modest. Compared with earlier candidate measures, methylation performs well. Leukocyte telomere length, discussed under [[telomeres-and-telomerase]], correlates with age so loosely in individuals that it is close to useless as a personal readout. Methylation clocks are far tighter, which is why they displaced telomere length as the field's default molecular measure of the [[hallmarks-of-aging|epigenetic hallmark]] within a few years.Compared with earlier candidate measures, methylation performs well. Leukocyte telomere length, discussed under [[telomeres-and-telomerase]], correlates with age so loosely in individuals that it is close to useless as a personal readout. Methylation clocks track age far more tightly, which is why within a few years they displaced telomere length as the field's default molecular measure of the [[hallmarks-of-aging|epigenetic hallmark]]. ## Generations of clocks## Generations of clocks lines 56–70 → 56–7012 unchanged lines not shown
### First generation### First generation Early clocks were trained directly on chronological age. [[steve-horvath]]'s 2013 multi-tissue clock used 353 CpG sites and predicted age across most human tissue types with a median error of roughly three and a half years, a result that also held for cells in culture and, notably, returned close to zero for [[induced-pluripotent-stem-cells]].[^horvath2013] Gregory Hannum's group published a blood-specific clock in the same year.[^hannum2013] Because these models are optimised to reproduce the calendar, any deviation from it is treated as error by the fitting procedure, which limits how much aging-relevant signal they can retain.Early clocks were trained directly on chronological age. [[steve-horvath]]'s 2013 multi-tissue clock used 353 CpG sites and predicted age across most human tissue types with a median error of roughly three and a half years, a result that also held for cells in culture. It returned an age close to zero for [[induced-pluripotent-stem-cells]].[^horvath2013] Gregory Hannum's group published a blood-specific clock in the same year.[^hannum2013] Because these models are optimised to reproduce the calendar, the fitting procedure treats any deviation from it as error, which limits how much aging-relevant signal they can retain. ### Second generation### Second generation The second wave changed the training target. DNAm PhenoAge was fitted to a composite of clinical laboratory measures and chronological age rather than to age alone.[^levine2018] GrimAge went further, building methylation surrogates for plasma proteins and for smoking history and then regressing those on time to death.[^lu2019] Both predict mortality and disease incidence considerably better than first-generation clocks, which is unsurprising given what they were trained on.The second wave changed the training target. DNAm PhenoAge was fitted to a composite of clinical laboratory measures and chronological age rather than to age alone.[^levine2018] GrimAge went further, building methylation surrogates for plasma proteins and for smoking history and then regressing those on time to death.[^lu2019] Both predict mortality and disease incidence considerably better than first-generation clocks, as might be expected from what they were trained on. ### Pace-of-aging### Pace-of-aging DunedinPACE takes a different approach again. Rather than estimating a level, it estimates a rate, trained on two decades of repeated organ-function measurements in the Dunedin birth cohort, all of whom are the same age.[^belsky2022] Its output is a speed: roughly one biological year per calendar year for an average person. Because chronological age is constant within the training set, the model cannot be fitting the calendar.DunedinPACE estimates a rate instead of a level. It was trained on two decades of repeated organ-function measurements in the Dunedin birth cohort, all of whom are the same age.[^belsky2022] Its output is a speed: roughly one biological year per calendar year for an average person. Because chronological age is constant within the training set, the model cannot be fitting the calendar. ## What the epidemiology shows## What the epidemiology shows lines 74–95 → 74–953 unchanged lines not shown
## Reliability and confounding## Reliability and confounding Technical noise is a serious and underappreciated problem. Running the same DNA sample twice on the same array platform can produce clock estimates differing by more than a year, which is large relative to the effect sizes reported in intervention studies. Reconstructing clocks on principal components of the methylation data substantially improves test-retest reliability.[^higgins2022]Technical noise is a serious and often overlooked problem. Running the same DNA sample twice on the same array platform can produce clock estimates differing by more than a year, which is large relative to the effect sizes reported in intervention studies. Reconstructing clocks on principal components of the methylation data substantially improves test-retest reliability.[^higgins2022] Laboratory precision is not the same as stability in the body. A 2026 comparison of 18 clocks found that blood drawn from the same people a few hours or days apart agreed far less well than repeat assays of one sample: a meal, acute psychological stress or air pollution was enough to shift readings, no clock reached the conventional threshold for excellent reliability, and how reproducible a clock was in the laboratory did not predict how stable it was across repeat samples.[^sehgal2026]A clock can be precise in the laboratory and still unstable in the body. A 2026 comparison of 18 clocks found that blood drawn from the same people a few hours or days apart agreed far less well than repeat assays of one sample: a meal, acute psychological stress or air pollution was enough to shift readings, no clock reached the conventional threshold for excellent reliability, and how reproducible a clock was in the laboratory did not predict how stable it was across repeat samples.[^sehgal2026] Cell composition is the other main confounder. Blood is a mixture, and the proportions of naive T cells, memory T cells, monocytes and granulocytes shift with age and with acute illness. A clock trained on whole blood partly measures that shift. Some clocks adjust for estimated cell counts; the adjustment is imperfect.Cell composition is the other main confounder. Blood is a mixture, and the proportions of naive T cells, memory T cells, monocytes and granulocytes shift with age and with acute illness. A clock trained on whole blood partly measures that shift. Some clocks adjust for estimated cell counts; the adjustment is imperfect. > [!caution] Contested> [!caution] Contested> No epigenetic clock has been shown to track a causal driver of aging rather than a downstream correlate. Methylation changes could be the mechanism, a readout of the mechanism, or an incidental consequence of cell-turnover history. The distinction is invisible to the regression.> No epigenetic clock has been shown to track a causal driver of aging rather than a downstream correlate. Methylation changes could be the mechanism, a readout of the mechanism, or an incidental consequence of cell-turnover history, and the regression cannot tell these apart. ## Response to interventions## Response to interventions The strongest use case for clocks would be as a short-cut endpoint, letting a trial read out in a year rather than in decades. See [[aging-biomarkers]] for the general problem and the regulatory bar. The evidence that clocks respond to interventions in humans is thin.The most valuable use for clocks would be as a short-cut endpoint that lets a trial read out in a year instead of decades; [[aging-biomarkers]] covers the general problem and the regulatory bar. The evidence that clocks respond to interventions in humans is thin. The most-cited human result is the TRIIM study, in which nine men received growth hormone with DHEA and [[metformin]] and showed a mean reduction of a few years in several clock readings alongside thymic changes.[^fahy2019] It had no control group and nine participants. In the CALERIE randomised trial of [[caloric-restriction]], stored samples showed a small slowing of DunedinPACE in the restricted arm but no significant effect on PhenoAge or GrimAge.[^waziry2023] That pattern, one clock moving and others not, recurs across studies and is not well explained.The most-cited human result is the TRIIM study, in which nine men received growth hormone with DHEA and [[metformin]] and showed a mean reduction of a few years in several clock readings alongside thymic changes.[^fahy2019] It had no control group and nine participants. In the CALERIE randomised trial of [[caloric-restriction]], stored samples showed a small slowing of DunedinPACE in the restricted arm but no significant effect on PhenoAge or GrimAge.[^waziry2023] That pattern, one clock moving and others not, recurs across studies and is not well explained. [[epigenetic-reprogramming]] moves clock readings dramatically in cells and in mice, which is the strongest demonstration that the readings are manipulable. It is also the clearest case where a clock delta and a functional benefit could come apart, since the same treatment that resets methylation can erode cell identity.[[epigenetic-reprogramming]] moves clock readings dramatically in cells and in mice, the strongest demonstration that the readings can be manipulated. It is also the clearest case in which a change in a clock and a functional benefit could come apart, since the treatment that resets methylation can also erode cell identity. ## Beyond humans## Beyond humans lines 98–110 → 98–1102 unchanged lines not shown
## Consumer testing## Consumer testing Several companies sell direct-to-consumer methylation age tests, through the same mail-in channels asSeveral companies sell direct-to-consumer methylation age tests, through the same mail-in channels asthe [[consumer-blood-testing|direct-to-consumer laboratory panels]]. The assays are usually real, but the reported precision often exceeds what the underlying reliability supports, the clocks used are frequently proprietary and unpublished, and a result expressed as a single number invites over-interpretation of what is a noisy estimate with wide individual uncertainty. Repeat testing of the same person over short intervals can produce swings that reflect assay variation or day-to-day biological fluctuation rather than any change in [[biological-age]]. As of October 2026 no consumer clock is an approved diagnostic test.the [[consumer-blood-testing|direct-to-consumer laboratory panels]]. The assays are usually genuine, but the reported precision often exceeds what the underlying reliability supports, and the clocks used are frequently proprietary and unpublished. A result given as a single number invites over-interpretation of a noisy estimate with wide individual uncertainty. Repeat testing of the same person over short intervals can produce swings that reflect assay variation or day-to-day biological fluctuation rather than any change in [[biological-age]]. As of October 2026 no consumer clock is an approved diagnostic test. The most publicised individual user is [[bryan-johnson]], who presents his DunedinPACE results as evidence for his [[blueprint-protocol|Blueprint]] regimen and runs a public leaderboard, the Rejuvenation Olympics, that ranks entrants by the same measure. In a 2025 post he reported a single reading of 0.48 and an average of 0.54 across three tests, described this as the world's slowest speed of aging, and said he chose the metric for "the robustness and reliability of the signal it measures".[^johnson2025] These are self-reported results for one person, with no control and a regimen that changes many things at once, and a lower reading is not by itself evidence that aging has slowed in function or in outcomes. DunedinPACE is highly reproducible when the same DNA is assayed twice, but the 2026 reliability study placed it among the less stable clocks across samples taken from the same person hours or days apart, and in the poor range across a laboratory stress test.[^sehgal2026] Daniel Belsky, who led its development and receives royalties from TruDiagnostic's consumer sales of the test, has said he would not restrict access to such tests "just because they're in beta testing", but that hearing of people basing health decisions on his tool makes him "a little anxious" and that he hopes users read results "with a grain of salt", since "any single result may have some error in it".[^zimmer2025]The most publicised individual user is [[bryan-johnson]], who presents his DunedinPACE results as evidence for his [[blueprint-protocol|Blueprint]] regimen and runs a public leaderboard, the Rejuvenation Olympics, that ranks entrants by the same measure. In a 2025 post he reported a single reading of 0.48 and an average of 0.54 across three tests, described this as the world's slowest speed of aging, and said he chose the metric for "the robustness and reliability of the signal it measures".[^johnson2025] These are self-reported results for one person, with no control and a regimen that changes many things at once. A lower reading is not by itself evidence that aging has slowed in function or in outcomes. DunedinPACE is highly reproducible when the same DNA is assayed twice, but the 2026 reliability study placed it among the less stable clocks across samples taken from the same person hours or days apart, and in the poor range across a laboratory stress test.[^sehgal2026] Daniel Belsky, who led its development and receives royalties from TruDiagnostic's consumer sales of the test, has said he would not restrict access to such tests "just because they're in beta testing", but that hearing of people basing health decisions on his tool makes him "a little anxious" and that he hopes users read results "with a grain of salt", since "any single result may have some error in it".[^zimmer2025] ## Outlook## Outlook Clocks are firmly established as research instruments and firmly unestablished as endpoints. The work that would change that is unglamorous: standardised assays, published reliability statistics, replication of intervention effects across independent cohorts, and demonstrations that a clock change predicts a later functional change in the same person. Competitions such as [[xprize-healthspan]] have deliberately chosen functional endpoints instead, on the view that measured muscle, cognitive and immune restoration is what the [[geroscience-hypothesis|geroscience]] argument ultimately has to deliver.Clocks are well established as research instruments and not established at all as endpoints. The work that would change that is unglamorous: standardised assays, published reliability statistics, replication of intervention effects across independent cohorts, and demonstrations that a clock change predicts a later functional change in the same person. Competitions such as [[xprize-healthspan]] have deliberately chosen functional endpoints instead, on the view that measured muscle, cognitive and immune restoration is what the [[geroscience-hypothesis|geroscience]] argument ultimately has to deliver. ## See also## See also removed, struck through added, underlinedLine numbers count the serialised markdown of each revision, frontmatter included.
117 lines are the same in both revisions, and 61 of them are not shown. Neither revision here is necessarily the current article — read Epigenetic clocks for the text this wiki stands behind today.