# A Machine‐Learning Model of Chronological Age Based on Routine Blood Biomarkers in a Central European Population: A Potential Biological Age Marker

## TL;DR

This study uses machine learning and blood biomarkers to estimate a person's age and suggests it could help measure biological aging in a Central European population.

## Contribution

A novel machine-learning model using routine blood biomarkers to estimate chronological age in a large Central European population.

## Key findings

- XGBoost achieved the best performance with an MAE of 8.73 years.
- Key biomarkers include ALT, creatinine, ALP, glucose, and MCV.
- The model uses accessible blood markers to estimate age but needs further validation against clinical outcomes.

## Abstract

Aging is a gradual decline in physiological and functional capacities that leads to an exponentially increasing risk of death. Although aging is universal, the rate of aging differs substantially between individuals. Biomarkers of aging are being developed to improve the prediction of a person’s susceptibility to disease onset, disease course, and complications, as well as to estimate lifespan and healthspan.

The primary aim of this study was to develop and evaluate machine‐learning models that estimate chronological age from routinely measured blood biomarkers in a large Central European population. A secondary aim was to characterize the relative contribution of individual biomarkers and to discuss the resulting index as a potential biological age marker.

We modeled chronological age as a regression problem using four algorithms: a multilayer neural network, Extreme Gradient Boosting (XGBoost), Random Forest, and Ridge Regression. The dataset comprised more than 26 million anonymized laboratory results from over 3 million individuals. Model performance was assessed using mean absolute error (MAE), root mean squared error (RMSE), mean absolute percentage error (MAPE), and epsilon‐accuracy. We also examined feature importance to identify the most informative biomarkers.

XGBoost achieved the best performance, with an MAE of 8.73 years across all ages. The 10 most influential predictors were alanine aminotransferase (ALT), creatinine, alkaline phosphatase (ALP), glucose, mean corpuscular volume (MCV), thrombocytes, albumin, mean corpuscular hemoglobin (MCH), urea, and aspartate aminotransferase (AST). These markers span hepatic, renal, metabolic, and hematological domains.

Using easily accessible blood biomarkers, it is possible to estimate chronological age with an MAE of 8.73 years in a large Central European population. Because the present work does not include validation against clinical outcomes, the resulting index should be regarded as a potential biological age marker. Future studies are needed to test its association with morbidity, mortality, and established biological age measures in independent cohorts.