Machine learning predicts biological age from blood metabolites

This video summarizes the paper “Metabolomic age (MileAge) predicts health and life span: A comparison of multiple machine learning algorithms” by Julian Mutz, Raquel Iniesta, and Cathryn M. Lewis (Science Advances, 2024; PMID 39693428). The study benchmarked 17 machine learning algorithms using nuclear magnetic resonance (NMR) metabolomics data from over 100,000 UK Biobank participants to build biological ageing clocks. The resulting MileAge model accurately predicted chronological age and was strongly associated with frailty, telomere length, chronic illness, and mortality. Individuals with an older metabolomic age showed worse health outcomes and shorter lifespan, suggesting that metabolite-based clocks can serve as scalable biomarkers for biological ageing and health risk assessment.