Precise and interpretable neural networks reveal epigenetic signatures of aging in youth across health and disease

DNA methylation age clocks are powerful tools for measuring biological age, providing insights into aging risks and outcomes beyond chronological age. While traditional clocks are effective, their interpretability is often limited due to their dependence on small, potentially stochastic sets of CpG sites. We propose that the reliability of epigenetic age estimation models should stem from their ability to detect comprehensive and targeted aging signatures. Here we introduce NCAE-CombClock, a neural network regressor that improves age prediction accuracy in large datasets by integrating methylation embeddings with CpG sites. Additionally, we developed explainable neural networks for robust age classification across adolescence and young adulthood. Epigenetic aging signatures identified at single-year resolution from these models were enriched in developmental, immune, and metabolic processes. We showcase the utility of our approach by exploring the biological mechanisms underlying the altered pace of aging observed in pediatric Crohn's disease. Our models offer broad applications in personalized medicine and aging research, providing a valuable resource for interpreting aging trajectories in health and disease.