For most of human history, we had no way to measure how fast a person was aging. Chronological age — the number of years since birth — was the best we had, and it is a deeply imperfect proxy. Today, using patterns of chemical modifications to DNA, scientists can estimate a person's biological age with remarkable precision. These tools — epigenetic clocks — are transforming longevity medicine.
The idea that two people of the same chronological age can have dramatically different biological ages is intuitive — we all know 60-year-olds who seem decades younger than their peers, and 40-year-olds who seem prematurely aged. What epigenetic clocks have provided is the first rigorous molecular basis for this observation: a quantitative, reproducible measurement of biological age that predicts outcomes far more accurately than the calendar alone.
DNA methylation is the addition of a methyl group (CH₃) to cytosine bases in DNA — most commonly at CpG dinucleotides (sites where cytosine is followed by guanine). These methylation marks do not change the genetic code, but they profoundly influence gene expression: methylated promoter regions typically silence the associated gene, while unmethylated regions are associated with active transcription.[1]
What makes methylation so useful for aging measurement is that it changes with age in highly systematic, predictable ways — some sites gaining methylation with age, others losing it — across all individuals, regardless of genetics. Steve Horvath at UCLA recognized that these age-correlated changes could be used algorithmically: by measuring methylation at carefully selected CpG sites and applying a regression model, you could predict chronological age from a blood or tissue sample with a median error of less than 3.6 years.[2] This was Horvath's landmark 2013 discovery — a molecular clock encoded in the epigenome itself.
Horvath's first-generation clock uses 353 CpG sites to predict chronological age across virtually every human tissue type — blood, saliva, brain, liver, skin, and more. It was the proof of concept that methylation-based age prediction was possible and reproducible. Its key limitation is that it predicts chronological age — not biological age per se, and not disease risk or mortality independently of chronological age. A person whose Horvath clock says 45 when they are chronologically 50 is biologically younger, but the clock does not directly tell you how much healthier they are.[3]
Developed by Morgan Levine at UCLA (now at Yale), PhenoAge represents a major advance over the original Horvath clock. Levine trained her model not on chronological age but on "phenotypic age" — a composite of nine clinical biomarkers (albumin, creatinine, glucose, CRP, lymphocyte percentage, MCV, RDW, alkaline phosphatase, and white blood cell count) that together predict disease risk and mortality. The resulting methylation clock therefore predicts biological phenotype rather than calendar age.[4] People whose PhenoAge exceeds their chronological age have accelerated biological aging and meaningfully elevated disease risk.
Developed by Horvath and colleagues, GrimAge is the most clinically powerful mortality predictor among current clocks. It was trained directly on time-to-death data and smoking pack-years — making it uniquely sensitive to the biological consequences of the most important longevity risk factors. In multiple prospective cohort studies, GrimAge acceleration (biological age ahead of chronological age) has been associated with substantially elevated all-cause mortality, cardiovascular disease risk, and cancer incidence.[5] If you can only use one clock to predict longevity outcomes, GrimAge has the strongest current evidence base.
DunedinPACE represents a conceptually distinct innovation — rather than estimating biological age at a point in time, it measures the current rate at which a person is aging. Developed from the Dunedin birth cohort, the clock outputs a number close to 1.0 in the average person, with values above 1.0 indicating faster-than-average aging and values below 1.0 indicating slower aging. This makes DunedinPACE particularly useful for measuring the effect of interventions on aging speed over time — and emerging data suggests it may be the most sensitive clock for detecting lifestyle intervention effects.[6]
"GrimAge tells you how old your biology has become. DunedinPACE tells you how fast you are getting there. Both questions are important, and they are not the same question."
— Daniel Belsky, PhD, Columbia University, lead developer of DunedinPACEThe most practically important question about epigenetic clocks is whether the biological age they measure can be altered — and if so, what alters it. The evidence is now reasonably clear: yes, epigenetic age is modifiable, and several lifestyle interventions have documented effects.[7]
The CALERIE trial — a rigorous multi-site RCT testing 25% caloric restriction in healthy non-obese adults — found that caloric restriction slowed the DunedinPACE rate of aging over two years, providing the first causal evidence from a controlled human trial that epigenetic aging pace is responsive to dietary intervention.[8]
Exercise has been associated with lower PhenoAge and GrimAge acceleration in multiple observational studies. The effect is dose-dependent and appears to be driven primarily by the anti-inflammatory and metabolic effects of regular physical activity rather than any single molecular pathway.
Diet quality — assessed via Mediterranean diet adherence, dietary inflammatory index, and overall dietary pattern — is consistently associated with lower epigenetic age in large cohort studies. The effect sizes are modest but real and consistent across multiple clock types and populations.[9]
Sleep matters more than often appreciated. Short sleep duration and poor sleep quality are associated with epigenetic age acceleration, with effects comparable to those of smoking in some studies. The mechanism likely involves disrupted DNA repair and inflammatory pathway activation during sleep-deprived states.
The consumer epigenetic testing market has grown rapidly, with products from TruAge (which uses GrimAge and DunedinPACE), Elysium Health's Index (Horvath-based), MyDNAge, and others now available directly to consumers at $200–$500 per test.[10]
Before purchasing, understand these key limitations: First, the test-retest reproducibility of consumer clocks is imperfect — the same person tested twice within weeks can show differences of 2–5 years, partly from biological variation and partly from laboratory variation. This means single-point measurements should be interpreted cautiously; the more meaningful use is serial testing (yearly) to track trends. Second, not all clocks are equivalent — a "biological age" from an Horvath-trained product tells you something different from a GrimAge or DunedinPACE result, and comparing across different test providers is difficult. Third, the clinical actionability of these tests, while improving, remains limited — there is no established medical protocol for responding to an elevated epigenetic age beyond the lifestyle recommendations that are beneficial regardless of your score.