Discover how a new metabolomic aging score outshines conventional metrics in predicting short-term mortality, opening doors to personalized health insights and early disease intervention.
Study: A metabolomic profile of biological aging in 250,341 individuals from the UK Biobank. Image Credit: ArtemisDiana / Shutterstock
In a recent study published in the journal Nature Communications, researchers from China investigated nuclear magnetic resonance (NMR) biomarkers associated with aging. They developed a longitudinal metabolomic aging rate and a metabolomic aging score to predict the risk for disease and all-cause mortality. They identified 54 representative aging-related biomarkers with varying hazard ratios, including GlycA, which had the highest hazard ratio (1.25 per SD) for all-cause mortality. The study also uncovered 439 potential causal biomarker-disease pairs through multivariable Mendelian randomization and colocalization analysis, leading to the creation of a metabolomic aging score that better predicts short-term mortality risk.
Background
Aging is a complex biological process that leads to declining physiological functions and increases the risk of frailty, disease, and mortality. In 2017, aging-related conditions contributed to over half of the global health burden in adults. The advancements in omics technologies have accelerated the research on biological aging, leading to the development of aging clocks that predict both chronological age and adverse health outcomes. This study highlights the utility of metabolomics, particularly through innovations in high-throughput NMR analysis and machine learning, for population-scale aging research and disease prediction. The United Kingdom (UK) Biobank's comprehensive NMR metabolomics data and health-related information is a crucial resource for advancing metabolomics-based aging research. In the present study, researchers investigated the aging-associated biomarkers and examined their predictive power for mortality. They additionally developed a novel metabolomic aging score and derived a personalized metabolomic aging rate.
About the study
The UK Biobank dataset included 249 metabolomic biomarkers (168 in absolute concentrations and 81 derived ratios) from about 250,341 participants. An additional 76 biomarker ratios were computed to supplement existing data, and quality control measures were employed. A LASSO Cox proportional hazards model was used to identify aging-related biomarkers.
A GWAS of 325 biomarkers was conducted on a subset of around 95,000 individuals to identify genetic variants associated with the biomarkers. Genetic correlations and pleiotropic effects were analyzed, and multiple aging metrics (e.g., frailty index, leukocyte telomere length) were compared to the metabolomic aging score. A multivariable Mendelian randomization (MVMR) analysis assessed potential causal relationships between metabolomic biomarkers and 20 aging-related diseases. Various statistical methods (e.g., MVMR-IVW, MVMR-Egger) were used, and colocalization analysis explored shared genetic variants between biomarkers and diseases.
Results and discussion
The study identified 54 metabolomic biomarkers linked to biological aging, including amino acids, ketone bodies, fatty acids, lipoproteins, and inflammation-related markers. GlycA, a systemic inflammation biomarker, showed the highest hazard ratio (1.25 per SD) for all-cause mortality. Most biomarkers correlated significantly with various aging metrics, such as chronological age, the frailty index, and leukocyte telomere length. While GlycA was linked to a higher likelihood of frailty, three polyunsaturated fatty acid biomarkers were associated with lower odds of frailty. Additionally, certain lipoprotein-related biomarkers showed negative associations with cardiovascular diseases. A total of 439 candidate causal relationships were identified between 213 NMR biomarkers and 20 aging-related diseases, with 14 pairs reaching Bonferroni-corrected significance. Chronic kidney disease (CKD) had the most candidate biomarkers. Key biomarkers linked to diseases included glucose for type 2 diabetes (T2D) and creatinine for CKD. Some biomarkers served as shared risk or protective factors across multiple conditions, while colocalization analysis revealed pleiotropic variants influencing various biomarkers and diseases.
Further, a novel metabolomic aging score was developed based on 54 representative NMR biomarkers, highly correlated with MetaboHealth and moderately with chronological age and the frailty index. It demonstrated strong predictive performance for all-cause mortality across follow-up intervals, especially in the 51–60 age group, where it significantly outperformed chronological age. The score was the most accurate among aging metrics, particularly for short-term mortality risk, while showing comparable performance to chronological age for 10-year prediction but less effectiveness for 15-year prediction. The study also developed a metabolomic aging rate, derived from longitudinal data, offering a more personalized assessment of aging. The metabolomic aging score effectively predicted disease risk, particularly for conditions with dysregulated metabolic pathways, outperforming other aging metrics for diseases like T2D and CKD. Differences in the metabolomic aging score distinguished between future early-onset, other-onset, and disease-free groups, with significant findings for diseases like T2D and hypertension. Additionally, 40 pro-aging and anti-aging biomarkers were identified, showing distinct patterns based on the metabolomic aging rate.
Although strengthened by its large scale, the study is limited by a narrow age range of participants, underrepresentation of disadvantaged groups, and the potential variability of the plasma metabolome's predictive power across different diseases. Interpretation of causal relationships in the study also warrants caution.
Conclusion
In conclusion, the present study offers the most comprehensive metabolomic profile linked to biological aging. It introduces a metabolomic aging score that can predict short-term mortality and disease risk, outperforming other aging metrics in specific contexts. However, the score is not intended as a definitive measure of biological aging. Instead, it reflects the aging signal at the metabolome level. Future research could potentially combine this score with other aging metrics, such as proteomic and epigenetic data, to further improve our understanding of aging.
Key risk factors influencing mortality rates in chronic pulmonary aspergillosis patients