Preprint models biological ageing as an information-theoretic problem using DNA methylation entropy

A preprint on bioRxiv describes a new approach to epigenetic age estimation that combines information theory with nonlinear machine learning, positioning DNA methylation entropy — the degree of variability in methylation patterns across a cell population — as a biologically meaningful measure of ageing.

Epigenetic clocks based on DNA methylation at CpG sites are among the most widely used molecular correlates of chronological and biological age. Most current clocks, however, are empirical models: they identify methylation sites that track with age and combine them into a weighted predictor, without explicitly modelling why methylation variability changes over time.

The authors borrow from statistical physics the concept of entropy — broadly, the spread or disorder of a distribution — and apply it to the population-level distribution of methylation beta values at individual CpG sites. They argue that age-related increases in methylation entropy reflect an erosion of epigenetic fidelity, analogous to physical processes governing disorder in other complex systems. The resulting predictor is described as both competitive in accuracy with established clocks and more interpretable, in that its components correspond to physically motivated quantities rather than empirical weightings.

The work contributes to a growing field using physics-inspired frameworks to understand ageing at the molecular level. It has not yet undergone peer review, and independent validation on diverse cohorts will be needed to assess generalisability.