A revised multi-tissue, multi-platform epigenetic clock model for methylation array data

Abstract Epigenetic changes have long been investigated in association with the process of aging humans. DNA methylation has extensively used as a surrogate measure biological age and correlations between “DNA age” chronological established. A wide variety epigenetic clocks designed to predict different tissues on data obtained from platforms. We aimed extend scope one most predictors, Horvath pan-tissue clock, improve its accuracy acquired latest Illumina platform (BeadChip EPIC). present three models trained close 6,000 samples various source platforms demonstrate their superior performance (Pearson correlation ( r ) = 0.917–0.921 median absolute error (MAE) 3.60–3.85 years) compared original model 0.880 MAE 5.13 test set more than 4,000 samples. The gain was especially pronounced EPIC array 0.89, 3.54 years vs. 0.83, 6.09 years), which not available at time when created. Our updated great precision an independent cohort multiple tissue types Two presented exclusively use covariates albeit coefficients, allowing for straightforward adaptation prefiltered datasets previously processed predictor.