DNA Methylation Clocks: What Lab Aging Tests Actually Measure

Two people born in the same year can be biologically a decade apart. That gap is not a metaphor — it is measurable in DNA. A methylation clock reads chemical tags attached to your genome and converts them into an age estimate that predicts mortality better than your birth certificate does. The best-known version, GrimAge, outperforms chronological age at forecasting time-to-death in large cohort studies, and the gap between your methylation age and your real age is now used as an outcome measure in clinical trials of exercise, diet and drugs.

If you want the simpler version of this comparison — how your body’s function tracks against calendar years — our biological age calculator gives you a rough estimate from everyday markers. What follows is what the lab-based tests actually measure and how to read the number when it comes back.

What a methylation clock actually reads

DNA methylation is the addition of a methyl group to cytosine bases, usually where a cytosine sits next to a guanine (a CpG site). It does not change the genetic code; it changes whether genes are switched on. Across the genome there are roughly 28 million CpG sites, and a few hundred of them change methylation status predictably with age.

A clock is a regression model trained on those sites. It takes a saliva or blood sample, measures methylation at specific CpGs, and outputs an age in years. The training set is the important part: a model trained on chronological age learns to predict time since birth, while a model trained on mortality data learns to predict something closer to biological wear.

Clock	Introduced	Trained on	Typical error vs. real age	Best use
Horvath (pan-tissue)	2013	Chronological age, 51 tissues	±3.6 years	Broad tissue comparison
Hannum	2013	Chronological age, blood	±4.9 years	Blood-based cohort work
PhenoAge	2018	Age plus clinical biomarkers	±6 years	Mortality and morbidity risk
GrimAge	2019	Mortality, smoking pack-years, plasma proteins	±5 years	Best mortality prediction
DunedinPACE	2022	Rate of decline in 19 organ systems	Not an age — a pace	Measuring speed of ageing

Why the error bars are the whole story

A clock that predicts age with a 3.6-year average error is not a precise instrument at the individual level. Two samples from the same person a month apart can differ by a year or more depending on the tissue, the array batch and even the time of day. Commercial kits often report a single number without a confidence interval, which makes a reading of “42.1 years” look far more definitive than the underlying measurement supports.

The meaningful signal is the trend, not the snapshot. A reading three years older than your calendar age is unremarkable noise. A reading that moves five years in the same direction across two annual tests is a real signal. This is why the research consensus favours repeated measurement over one-off testing.

What moves a clock reading in 12 months

Several randomised and quasi-experimental trials have shown measurable clock movement over periods short enough to be motivating. The effect sizes are modest but consistent.

• Caloric restriction: the CALERIE trial found roughly a 2 to 3 year slowing of PhenoAge and GrimAge progression over two years in the restricted group versus control.
• Exercise volume: higher cardiorespiratory fitness and greater weekly vigorous activity correlate with younger GrimAge, with the strongest effects in people crossing from sedentary to moderately active.
• Smoking cessation: GrimAge explicitly includes pack-years, so quitting visibly decelerates the clock — one of the largest single-lever effects available.
• Sleep regularity: shift workers and people with highly variable sleep timing show accelerated methylation ageing in cohort data, independent of total sleep duration.

The shift-work finding is worth pausing on

Among the more striking results in this literature is the association between long-term night shift work and accelerated epigenetic ageing. A study of nurses with years of rotating night shifts found measurably older methylation age relative to calendar age, with the gap scaling roughly with duration of shift exposure. The plausible mechanism is chronic circadian misalignment, which disrupts the methylation patterns that clocks read in the first place.

This does not mean a single year of late shifts costs you years of life. The effect sizes are measured in fractions of a year per year of exposure. It does mean that job-level circadian disruption belongs in the same conversation as smoking and sedentary behaviour when you are trying to explain a clock reading that looks older than expected.

How to interpret a commercial test result

If a kit reports that you are 4.2 years older than your birth age, the honest reading is: you are within the normal error band, and one data point tells you almost nothing. Note the tissue used (saliva and blood are not interchangeable), record the exact clock version, and plan a second test at the same time of year under similar conditions. Ask whether the reported number is a raw clock estimate or a mortality-adjusted one, because those are different quantities with different reference populations.

What the clocks are genuinely good for right now is research and longitudinal comparison. As a personal dashboard, they are early. As a way to compare interventions in a population, they are already one of the more responsive biomarkers available. Tracking your functional age alongside these lab numbers is the more useful combination — you can start with the free estimate at YourAgeCalculator and use lab results to refine the picture over time.