The Epigenetic Pacemaker is a more sensitive tool than penalized regression for identifying moderators of epigenetic aging

Abstract

Epigenetic clocks, DNA methylation based chronological age prediction models, are commonly employed to study age related biology. The error between the predicted and observed age is often interpreted as a form of biological age acceleration and many studies have measured the impact of environmental and other factors on epigenetic age. Epigenetic clocks are fit using approaches that minimize the error between the predicted and observed chronological age and as a result they reduce the impact of factors that may moderate the relationship between actual and epigenetic age. Here we compare the standard methods used to construct epigenetic clocks to an evolutionary framework of epigenetic aging, the epigenetic pacemaker (EPM) that directly models DNA methylation as a function of a time dependent epigenetic state. We show that the EPM is more sensitive than epigenetic clocks for the detection of factors that moderate the relationship between actual age and epigenetic state (ie epigenetic age). Specifically, we show that the EPM is more sensitive at detecting sex and cell type effects in a large aggregate data set and in an example case study is more sensitive sensitive at detecting age related methylation changes associated with polybrominated biphenyl exposure. Thus we find that the pacemaker provides a more robust framework for the study of factors that impact epigenetic age acceleration than traditional clocks based on linear regression models.

Pseudotime Analysis Reveals Exponential Trends in DNA Methylation Aging with Mortality Associated Timescales

“…By construction, the age estimates produced by this approach (DNAm age) are constrained to be linear with chronological age [ 8 , 9 , 21 , 22 ]. Some epigenetic clocks, such as Horvath’s clock, address this limitation by modeling the methylation dynamics of children with a nonlinear function of chronological age [ 10 , 25 ]. However, the methylation aging trend that they captured may be biased due to a priori assumptions about the functional form underlying the relationship between methylation and chronological age.…”

“…Since epigenetic clocks are typically fit by minimizing the difference between observed and predicted ages (DNAm age), such an approach enforces linearity between the two measures [ 8 , 9 , 21 , 22 ]. Although some epigenetic clocks, such as Horvath’s clock, attempt to model age-associated methylation changes in children as a specific nonlinear function of time [ 10 , 23 ], the nonlinear trends captured by these clocks may be biased due to a priori assumptions about the functional relationship between epigenetic changes and age [ 24 , 25 ].…”