Could a simple blood test reveal how well someone is aging? A team of researchers led by Wolfram Weckwerth from the University of Vienna, Austria, and Nankai University, China, has combined advanced metabolomics with cutting-edge machine learning and a novel network modeling tool to uncover the key molecular processes underlying active aging. Their study, published in the Nature Journal npj Systems Biology and Applications, identifies aspartate as a dominant biomarker of physical fitness and maps the dynamic interactions that support healthier aging.
It has long been known that exercise protects mobility and lowers the risk of chronic disease. Yet the precise molecular processes that translate physical activity into healthier aging remained poorly understood. The researchers set out to answer a simple but powerful question: Can we see the benefits of an active lifestyle in elderly individuals directly in the blood – and pinpoint the molecules that matter most?
From fitness tests to blood fingerprints: A Body Activity Index and a Metabolomics Index
Researchers first synthesized a single "Body Activity Index" (BAI) by applying canonical correlation analysis to scores from walking distance, chair‐rise tests, handgrip strength, and balance assessments. This composite physical‐performance metric captures endurance, strength, and coordination in one robust measure. Independently, a "Metabolomics Index" was derived from blood concentrations of 35 small-molecule metabolites. Across 263 samples from older adults, these two indices showed a Pearson correlation coefficient of 0.85 (p < 1 × 10⁻¹⁹), demonstrating that the molecular signature in blood mirrors the composite measure of physical fitness.
Machine learning highlights active and less-active groups and their metabolic signature
To capture complex, non-linear patterns, the researchers trained five different machine-learning models – ranging from simple statistical approaches (Generalized Linear Model, GLM) to more advanced methods such as boosted decision trees (Gradient Boosting Machine, GBM; XGBoost) and a deep-learning autoencoder network. Each model was tuned with repeated cross-checks (double cross-validation) and tested on independent data to ensure robust performance. Both boosting methods (GBM and XGBoost) achieved high accuracy, distinguishing 'active' from 'less-active' participants in over 91% of cases (area under the curve, AUC > 0.91). Across all five algorithms, eight metabolites consistently emerged as predictors of activity level: aspartate, proline, fructose, malic acid, pyruvate, valine, citrate, and ornithine. Among them, aspartate stood out by a factor of two to three, confirming its central role as a molecular marker of active aging.
Network rewiring revealed by COVRECON
Correlation alone cannot explain why certain molecules are linked to fitness. To uncover the underlying mechanisms, the team applied COVRECON, a data-driven modeling tool. In simple terms, COVRECON looks at how metabolites vary together and then reconstructs the network of biochemical interactions between them. Mathematically, this involved estimating a differential Jacobian matrix – a way of identifying which enzymatic connections change most between active and less-active groups. This analysis revealed two well-known enzymes, aspartate aminotransferase (AST) and alanine aminotransferase (ALT), as central hubs in the network. Both are standard markers in clinical liver panels, but here they emerged as indicators of how activity reshapes metabolism. Importantly, the predictions were confirmed by routine blood tests: over the six-month study period, AST and ALT fluctuated more strongly in active participants than in their less-active peers – suggesting greater metabolic flexibility in liver and muscle function.
Implications for brain health and dementia
Aspartate is more than a simple metabolic intermediate: in the brain it also serves as a precursor of neurotransmitters, activating NMDA receptors that are essential for learning and memory. This dual role provides a possible link between physical fitness and cognitive health. Independent studies have shown that low AST and ALT levels in midlife – or an elevated AST/ALT ratio – are associated with increased risk of Alzheimer's disease and age-related cognitive decline. By demonstrating that physical activity drives dynamic changes in aspartate metabolism and in the plasticity of these two enzymes, the present study points to a molecular bridge between muscle-liver health and brain resilience. These findings suggest a simple message: physical activity helps in preserving strength and mobility, and may also contribute to protecting the brain from dementia through measurable shifts in amino-acid–based signaling pathways.
"Physical activity does more than building up muscle mass," explains Wolfram Weckwerth: "It rewires our metabolism at the molecular level. By decoding those changes, we can track – and even guide—how well someone is aging".
University Vienna Research Platforms which initiated this Project:
VIENNA METABOLOMICS CENTER.
Research Platform Active Ageing.
Original publication:
Jiahang Li, Martin Brenner, Iro Pierides, Barbara Wessner, Bernhard Franzke, Eva-Maria Strasser, Steffen Waldherr, Karl-Heinz Wagner & Wolfram Weckwerth. Machine learning and data-driven inverse modeling of metabolomics unveil key processes of active aging. In npj Systems Biology and Applications.
DOI: 10.1038/s41540-025-00580-4