Epigenetic Drift

Epigenetic drift refers to the gradual, largely stochastic accumulation of changes in DNA methylation marks across the genome as cells divide and age, distinct from the programmed methylation changes of development [1]. Some CpG sites become more methylated with age, others less, and the between-individual variance in methylation at many sites increases with age, meaning identical-twin methylomes that are similar at birth diverge over decades [1].

Steve Horvath's 2013 pan-tissue clock used methylation at 353 CpG sites across 51 tissue types to predict chronological age with a median error of about 3.6 years, establishing that epigenetic drift is regular enough to build a predictive clock despite being partly stochastic [2]. Second-generation clocks (PhenoAge, GrimAge) trained instead on mortality and clinical biomarkers rather than chronological age, and outperform Horvath's original clock at predicting time-to-death and disease onset [2].

What the evidence does not show: it remains unresolved how much of epigenetic drift is a cause versus a downstream marker of other aging processes, and a 2024 paper made the case that epigenetic clock predictions partly reflect drift-driven noise rather than a coordinated biological program, complicating interpretation of small clock changes after an intervention [1]. No RCT has established that reversing methylation drift at specific loci changes health outcomes in humans; partial epigenetic reprogramming has only been shown in animal models. Bryan Johnson cited a preprint reporting accelerated epigenetic aging across 16 different clocks in some rapamycin users as one factor in his decision to stop taking the drug, an example of clock-based readouts directly influencing a real-world self-experimentation decision rather than remaining a purely academic metric [4]. A separate methods-focused paper cautioned that different epigenetic clocks disagree on the direction of change after the same intervention often enough that clock selection alone can flip a reported conclusion, a recommendation to report results across multiple validated clocks rather than one [5].

Bryan Johnson's Blueprint protocol tracks a "Speed of Aging" DNA methylation-based metric; he has publicly reported a pace-of-aging value of 0.69 (implying he ages roughly 31% slower than the population average by that metric), an n=1 self-report from one clock, not independently replicated across labs [3].

Critics, including researchers who authored the 2024 drift paper, caution that epigenetic clocks can be sensitive to assay batch effects and that improvements on one clock after an intervention do not always replicate across other validated clocks, a concern that became concrete when a preprint reported rapamycin use was associated with accelerated aging on several of 16 tested clocks even as users expected the opposite [1]. A broader review of DNA-methylation-based biomarkers reached a similar conclusion about clock heterogeneity, noting that first- and second-generation clocks were built with different training targets and are not interchangeable measures of the same underlying biology [6].

The plain takeaway: epigenetic drift is a real, measurable, and partly predictable process, but translating small clock movements into confident claims about an individual's biological aging rate remains uncertain.