NMR Metabolomic Clocks Detect Accelerated Aging in Cancer and Liver Disease

Biological age—reflecting cumulative physiological decline from genetic, environmental, and lifestyle factors—often predicts health outcomes better than chronological age. Molecular aging clocks, trained on biomarkers from blood or tissue, have emerged as powerful tools to quantify this divergence. This perspective reviews the major clock types and introduces a new high-performance NMR-based metabolomic aging clock with direct clinical applications.

The authors situate their work within four major clock categories: epigenetic clocks (e.g., Horvath, Hannum) based on DNA methylation; transcriptomic clocks capturing gene regulatory shifts; proteomic clocks predicting multimorbidity and mortality; and metabolomic clocks detecting NAD+ depletion, mitochondrial dysfunction, and other aging-associated metabolic changes. Each approach has trade-offs in accuracy, interpretability, and biological insight. Prior NMR-based metabolomic models—including urine-based scores, the Estonian mortality biomarker study (17,345 individuals), the FINNRISK cohort (44,168 individuals), and the Dutch metaboAge model (18,000 serum samples)—showed moderate correlations with chronological age (Pearson ~0.65) and limited ability to resolve accelerated aging regions.

The CIC bioGUNE team built their clock on the AKRIBEA cohort of 13,500 Basque Country workers, supplemented to ~20,000 individuals for balanced age-spectrum coverage. A robust ensemble stacking machine learning approach applied to 1D ¹H-NMR NOESY spectra achieved a Pearson correlation of 0.92, with >75% of predictions within 10 years of chronological age—substantially outperforming prior models. A parallel interpretable version using quantified metabolites and clinical NMR parameters reached R=0.88, enabling mechanistic insight alongside predictive power.

Critically, the clock was applied to disease cohorts to assess metabolic distortion—the difference between metabolic and chronological age. In 717 prostate cancer patients (mean age ~67), metabolic age was accelerated by +4.9 ± 9.2 years (p=1.0×10⁻¹⁹). In 169 MASLD patients (mean age ~65), acceleration was more pronounced at +14.5 ± 10.9 years (p=1.9×10⁻²⁹), with greater heterogeneity. MASLD subtype A patients showed lower metabolic distortion than subtypes B+C, consistent with previously identified distinct serum lipidomic profiles, suggesting the clock can stratify disease severity and metabolic subtypes.

The authors acknowledge key limitations of all aging clocks: correlation does not imply causation; models trained on specific populations may not generalize across ethnicities or socioeconomic groups; single time-point measurements cannot capture aging dynamics; and the clinical impact of interventions that reduce biological age remains uncertain. They advocate for longitudinal multi-omic integration and population-diverse training sets as the field matures.