Serum metabolic signatures associated with maximal voluntary ventilation in native high-altitude Tibetans

This study aimed to identify serum metabolites independently associated with maximal voluntary ventilation (MVV) among native high‑altitude Tibetan adults, and explore metabolic correlates of pulmonary ventilatory function under chronic hypoxic exposure.

Untargeted serum metabolomic profiling comprising 5310 initially detected metabolic features, with 3381 high-quality annotated metabolites retained after strict quality filtering and comprehensive clinical data were collected from 168 native high‑altitude Tibetan adults with complete clinical, phenotypic, and metabolomic data (no missing values in the final analytical dataset). Continuous phenotypic and metabolomic variables were Z‑score‑standardized before analysis. Participants were stratified by MVV quartiles and randomly split into a training set (70%, n = 119) and an independent validation set (30%, n = 49). A multi‑tier metabolite‑screening pipeline was implemented solely within the training set: tiered‑threshold univariate regression generated candidate metabolites, followed by elastic‑net regularized regression (α = 0.5, lambda.1se). A prespecified fallback rule selected the top‑five metabolites with the smallest univariate P‑values when elastic‑net produced no non‑zero‑coefficient features. Selected candidates were entered into confounder‑adjusted multivariable linear regression, retaining metabolites with P < 0.05. Bootstrap resampling assessed screening stability. Ten‑fold cross‑validation evaluated internal performance, and the final model was tested on the independent validation set. Sensitivity analysis was performed by excluding the exogenous drug‑related metabolite to evaluate robustness of core associations.

Three serum metabolites showed independent associations with MVV: N‑Undecanoylglycine (β = -5.237, P = 0.002), Isosorbide Dinitrate (β = 0.245, P = 0.007), and Hypoglycin B (β = 4.655, P = 0.006). The metabolite‑based model yielded a 10‑fold cross‑validated R² of 0.191 in the training set, with predictive R² decreasing to 0.090 in the independent validation set. Bootstrap inclusion frequencies for target metabolites were moderate, suggesting relatively weak marginal metabolite‑MVV effects. Sensitivity analysis confirmed that core endogenous associations were not driven by the exogenous Isosorbide Dinitrate. Formal pathway enrichment was not conducted due to the small number of significant metabolites.

This study identified three serum metabolites independently correlated with MVV in native high‑altitude Tibetans using a multi‑stage metabolome‑wide screening workflow. These metabolites represent preliminary metabolic correlates of ventilatory function under chronic high‑altitude hypoxia. The marked attenuation of predictive performance from internal cross‑validation to external validation reflects limited generalisability of metabolic signatures in moderate‑sized population cohorts. This work emphasises the value of prespecified multi‑tier screening, stability evaluation and fallback strategies for metabolome‑wide association analyses.