Plasma metabolomic profiling across 18,476 individuals aged 1 to 96 years identified 56 age-associated metabolites and enabled development of metabolomic aging clocks associated with health status, chronic disease risk, and mortality.
Key Findings
Methods
A plasma metabolomic profiling approach using nano-assisted laser desorption/ionization mass spectrometry identified 56 representative age-associated metabolites across the human life span.
The study profiled plasma metabolomes from 18,476 participants aged 1 to 96 years across six centers in China.
The analytical platform used was nano-assisted laser desorption/ionization mass spectrometry (LDI-MS).
Participants spanned the full human life span from infancy to late old age.
Six centers contributed samples, providing geographic and demographic diversity within China.
Methods
A stacking ensemble model was developed to estimate metabolomic age in a healthy cohort, with metabolomic age acceleration (mAA) quantifying each individual's deviation from chronological age.
The model was built in a healthy cohort using the 56 identified age-associated metabolites.
Metabolomic age acceleration (mAA) was defined as each individual's deviation from their chronological age as predicted by the model.
The stacking ensemble approach combined multiple learning algorithms to improve prediction accuracy.
mAA served as a continuous individualized metric of biological aging relative to chronological age.
Results
Higher metabolomic age acceleration was associated with health deterioration and unfavorable lifestyles.
Individuals with higher mAA showed evidence of health deterioration across multiple measures.
Unfavorable lifestyle factors were linked to elevated mAA values.
The association between mAA and health status was observed within the healthy cohort used to develop the model.
The direction of association indicated that accelerated metabolomic aging corresponded with worse health profiles.
Results
Individuals with accelerated mAA showed increased risk of incident chronic diseases during follow-up from 2019 to 2025.
The prospective follow-up period extended from 2019 to 2025.
Incident chronic diseases associated with accelerated mAA included fatty liver disease, hypertension, and diabetes.
The association was prospective, meaning elevated mAA preceded the onset of these conditions.
This finding supports the predictive validity of the metabolomic aging clock for future disease risk.
Results
Accelerated metabolomic age acceleration was associated with increased all-cause mortality.
The mortality association was observed during the prospective follow-up period from 2019 to 2025.
All-cause mortality, not disease-specific mortality, was the outcome assessed.
This finding extends the clinical relevance of mAA beyond incident disease to survival outcomes.
The association was identified in the same cohort used for chronic disease risk analyses.
Results
A mortality-linked metabolomic metric called 'age deviation' was developed in a disease cohort to quantify individualized departures from age-expected mortality risk.
Age deviation was constructed from plasma metabolites to reflect individualized mortality risk relative to age-expected levels.
The metric was developed separately in a disease cohort, distinct from the healthy cohort used for mAA.
Age deviation was associated with established composite health measures.
Age deviation increased with multisystem disease involvement and higher multimorbidity burden, indicating its sensitivity to disease complexity.
Results
Age deviation increased with both multisystem disease involvement and higher multimorbidity burden.
As the number of organ systems affected increased, age deviation values rose correspondingly.
Higher multimorbidity burden—having multiple concurrent diseases—was associated with greater age deviation.
This dose-response pattern with disease complexity supports the metric's construct validity.
The findings were observed in the disease cohort portion of the study population.
Results
Plasma metabolomic profiling was demonstrated as a scalable, life span-wide approach for deriving clinically interpretable aging metrics.
The approach spanned participants aged 1 to 96 years, covering essentially the full human life span.
The LDI-MS platform enabled high-throughput profiling across 18,476 individuals at six centers.
Two distinct clinically interpretable metrics were derived: mAA for healthy populations and age deviation for disease populations.
The authors characterize the approach as 'scalable' and 'life span-wide,' suggesting potential for broad clinical application.
What This Means
This research suggests that a single blood test measuring dozens of small molecules (metabolites) in the bloodstream can reveal how quickly a person is biologically aging, regardless of their actual calendar age. The researchers analyzed blood samples from nearly 18,500 people in China ranging from infants to 96-year-olds and identified 56 metabolites that change predictably with age. Using advanced machine learning, they built a model that estimates a person's 'metabolomic age'—and then measured how much each person's biological age diverges from their chronological age. People whose biological age ran ahead of their calendar age (accelerated aging) were more likely to develop serious conditions like fatty liver disease, high blood pressure, and diabetes over the following years, and were also more likely to die from any cause.
For people who already had diseases, the researchers developed a related measure called 'age deviation' that captures how much a person's metabolic profile departs from what would be expected for someone their age with their mortality risk. This measure tracked closely with how sick people were overall—rising in step with the number of body systems affected and the number of diseases a person had simultaneously. Together, the two metrics offer a way to assess biological aging and health risk at the individual level across the entire human life span.
This research suggests that routine blood metabolite testing could eventually serve as a practical tool for identifying people at elevated risk of aging-related diseases before those diseases develop, potentially enabling earlier intervention. The study's large size and multi-center design strengthen confidence in the findings, though the approach was tested exclusively in Chinese populations, and further research would be needed to determine how well these specific metabolite patterns generalize to other populations.
Cao L, Chen Y, Zhang W, Jiang H, Huang H, Xu Y, et al.. (2026). Metabolomic aging and mortality clocks based on ensemble learning identify systemic aging risk across the human life span in 18,476 individuals.. Science translational medicine. https://doi.org/10.1126/scitranslmed.adz4815