Epigenetic Clocks: Measuring Biological Age and What It Means

Horvath's epigenetic clock (2013) was a genuine breakthrough — 353 CpG sites whose methylation patterns predict chronological age with a median error of 3.6 years across 51 different tissues. That's remarkable. But the deeper question is whether the clock measures something causally related to aging or whether it's a sophisticated correlation. The second-generation clocks (GrimAge, PhenoAge) were trained not on chronological age but on mortality and health outcomes, and they predict death significantly better than Horvath's original clock. GrimAge in particular shows hazard ratios of about 1.5-2.0 per year of age acceleration in multiple cohorts.

The causal question matters enormously for intervention. If epigenetic changes are downstream markers — the smoke, not the fire — then normalizing them won't help. But there's growing evidence they might be partly causal. Yamanaka factor-based reprogramming (partial dedifferentiation using OSKM or OSK) can reset epigenetic age in cells without fully dedifferentiation them, and the rejuvenated cells show functional improvements. The Sinclair lab's work on optic nerve regeneration in mice (Lu et al., 2020) using OSK factors was dramatic — old mice recovered youthful gene expression and regained vision after glaucoma-like injury. The work has been somewhat controversial, and the broader claims from that lab sometimes outrun the data, but the core finding on epigenetic reprogramming has been replicated.

What fascinates me is the information-theoretic angle. If aging is partly an accumulation of epigenetic noise — a loss of the information that tells cells what type they are and how to behave — then aging is fundamentally a signal processing problem. The backup copy of youthful epigenetic information might still exist somewhere (this is Sinclair's 'information theory of aging'), and reprogramming accesses it. I'm skeptical of the strongest version of this claim, but the framework is productive.

For a practical longevity intervention stack, epigenetic clocks give you something invaluable: a surrogate endpoint. You don't have to wait 30 years to see if an intervention extends lifespan. You can measure epigenetic age before and after and get a signal in months. This is what makes the Altos Labs bet on reprogramming viable as a research strategy, even if the final therapeutic applications are distant.