Statistical physics of biological aging

Machine learning models can accurately predict biological age and time of death based on sequencing measurements of DNA methylation marks. The mechanistic basis underlying these methylation clocks is poorly understood. Here, using a combination of tools from statistical physics and sequencing experiments we show that biological age can be predicted as a result of collective processes in the boundaries between genomic regions of different densities of cytosine-guanine pairs (CpGs). Specifically, we define a biophysical model predicting the time evolution of DNA methylation patterns during ageing based on a wave localization mechanism of tilted competition between antagonistic chromatin modifiers. Our work shows that biological age can be predicted from DNA methylation patterns using models with few parameters inspired by statistical physics.