DNA-methyaltion-based age predictor

DNA-methyaltion-based age predictor estimates chronological age from DNA methylation patterns (epigenetic clock) to enable study of biological aging and its association with mortality.

Key Features:

• Data-Driven Approach: Trained on 13,661 DNA methylation samples, including 13,402 blood and 259 saliva samples.
• Improved Prediction Accuracy: Prediction accuracy increases as training sample size grows from 335 to 12,710, approaching near-perfect chronological age estimates.
• Confounding Factors Correction: Predictive models incorporate correction for confounders such as cellular composition to reduce bias in age estimates.
• Association with Mortality: Evaluates the relationship between age acceleration residual (AAR) and mortality, reporting that the AAR–mortality association diminishes with improved prediction accuracy and that the best predictor showed no significant association in the Lothian Birth Cohorts of 1921 and 1936.
• Cross-Tissue Performance: Employs a multi-tissue-based approach and demonstrates robust performance in non-blood tissues in addition to blood.

Scientific Applications:

• Biological Aging Research: Provides precise chronological age estimates to investigate molecular signatures and biomarkers of aging.
• Epidemiological Studies: Enables analysis of lifespan-related associations by attenuating confounding in age acceleration residuals when examining mortality and population cohorts.
• Clinical Applications: Supplies quantitative biological age estimates that can be used as indicators in studies of disease risk and therapeutic outcomes.

Methodology:

Models were developed using Elastic Net regression (combining L1 and L2 regularization) with iterative refinement through large-scale data analysis and validation across diverse cohorts.