The chemical switch
The best-studied mark is methylation: a methyl group (one carbon, three hydrogens) placed on a cytosine in the DNA. The dot does not change the letter (it is still a C), but it silences the gene that carries it, because the reading proteins can no longer get in.
The exact picture: the same cookbook, with some pages stuck shut. The recipes exist; nobody reads them.
Why identical twins diverge
Identical twins share 100% of their DNA. At birth their epigenomes are nearly indistinguishable; over the decades they diverge measurably, as diet, smoking, stress and sleep leave their signature in methylation.
It is the demonstration that the environment writes on the genome without editing it. Frisén and colleagues documented it in PNAS (2005), comparing 3-year-old and 50-year-old twins.
And is it inherited? Here precision helps. From cell to cell, yes: when a neuron divides, most of its marks are copied, which is why a liver cell begets liver cells and not neurons. From parents to children, almost not: at fertilization the embryo wipes most marks from both gametes and reprograms the lot. A few methylation islands escape the wipe, the phenomenon called genomic imprinting, a few dozen confirmed regions. The line "you inherited your grandfather's stress" is, in humans, a hypothesis with indirect evidence, not an established fact.
The 2026 news
An epigenetic clock reads hundreds of CpG sites and estimates biological age. In August 2026 a Yale team harmonized 51 human intervention trials (3,128 samples) and recomputed 16 clocks under the same criteria, the largest review to date.
The result in one line: of 51 interventions, 19 reduced epigenetic age (13 after statistical correction), 5 increased it, and more than half moved no clock at all. Drugs and lifestyle did move needles; supplements, not consistently.
The practical lesson? Second-generation clocks (DunedinPACE, PCGrimAge), trained against mortality or the pace of aging, respond; first-generation ones, trained to guess your chronological age, barely. And no clock is yet validated as a clinical endpoint: buying one to "know your real age" buys a research estimate, not a diagnosis.
| Intervention | Trials | Clocks that moved | Reading |
|---|---|---|---|
| Lifestyle (diet, exercise, sleep) | 21 | 9 reduced epigenetic age (7 after correction) | Modest but measurable; the effect you can buy with habits |
| Drugs (metformin, rapamycin, etc.) | 14 | 6 reduced (4 after correction) | A real signal; dose and duration still to define |
| Supplements and combinations | 16 | 4 reduced (2 after correction); 5 increased | Inconsistent: the noisiest category and the least reliable |
Figures derived from the harmonized analysis of 51 trials (3,128 samples, 16 clocks). "After correction" = significance that survives multiple-testing adjustment.
Three limits the review itself underlines and no sales page mentions. First, no minimum clinically meaningful change is defined: if your clock drops 0.8 years, nobody knows whether that means anything for your health. Second, the clocks don't always agree with each other: in the same trial one can fall while another rises, because each algorithm captures a different signal. Third, the measured effects are small next to the variation between people: on the order of weeks or months of epigenetic age, not years of life.
Translation of the translation: epigenetics is already a real research tool and will likely become one of medicine. As a consumer product, today, a commercial clock is a number with poor instructions for use.