Highlights

• A Transformer-based foundation model was developed to quantify biological aging from longitudinal electronic health records, capturing health trajectories rather than single-time-point risk. 

• Differences between biological age and chronological age encode early, system-wide physiological decline and predict future disease risk with high accuracy across multiple conditions. 

• By modeling both pediatric development and adult aging, the study establishes biological age as a life-course metric for preventive and population health. 

Summary

This study presents a life-cycle biological clock that estimates biological age from routine clinical data using a Transformer-based architecture optimized for longitudinal electronic health records. Unlike conventional aging clocks or risk scores that rely on static snapshots, the model learns long-range temporal patterns in patient histories, creating a compact latent representation of an individual’s health trajectory across time. 

 

The scientific significance lies in how biological age is linked to health risk through age acceleration, defined as the divergence between biological and chronological age. This divergence reflects declining physiological reserve driven by cumulative changes in metabolism, inflammation, organ function, and immune regulation—processes that precede the onset of chronic disease. Using this framework, the model demonstrated strong predictive performance, achieving AUCs of approximately 0.97–0.98 for prevalent cardiometabolic diseases and 0.81–0.91 for 10-year incidence prediction across multiple conditions, outperforming traditional machine-learning baselines. 

 

A key biological advance is the discovery that aging is not a single continuous process. The model identifies two distinct life phases: a pediatric developmental clock and an adult aging clock, each governed by different biomarker patterns. This reframes biological age in children as a measure of physiological maturity, enabling quantitative assessment of accelerated or delayed development. Importantly, the model extracts predictive signals from routine laboratory values even within normal reference ranges, revealing early, individual-specific deviations before clinical thresholds are crossed. By combining large-scale data, methodological rigor, and low-cost clinical inputs, this work establishes biological age as a scalable, system-level metric for early risk stratification, preventive medicine, and population health management. 

K. Wang et al., “A full life cycle biological clock based on routine clinical data and its impact in health and diseases,” Nat. Med., 2025, doi: 10.1038/s41591-025-04006-w