Analytical workflows in salivary metabolomics for athlete stress monitoring: sampling, platform selection, and translation challenges

Abstract

Salivary metabolomics is increasingly positioned as a low-burden biochemical layer for athlete monitoring, but its translational value depends less on biomarker novelty alone than on whether the analytical workflow is sufficiently standardized for repeated real-world use. A methods-centered assessment of saliva therefore focuses on how collection control, preprocessing, extraction strategy, separation chemistry, platform selection, normalization, and quality assurance determine interpretability. Current athlete and exercise studies consistently indicate that salivary metabolic profiles respond to acute exercise, repeated competition, and training-related strain, with recurrent signals in amino acids, hydrophilic stress-related metabolites, and broader multimetabolite signatures. However, the literature remains limited by small cohorts, insufficient longitudinal validation, inconsistent sampling states, and incomplete reporting of front-end handling and analytical QA. Comparative reading across NMR, LC-MS, GC-MS, and CE-MS suggests that no single platform is universally optimal: NMR offers strong reproducibility and baseline phenotyping; LC-MS offers broad coverage but requires stronger quantitative discipline; GC-MS remains useful for derivatizable central metabolites; and CE-MS is especially informative for polar and ionic salivary metabolites, although its practical deployment is constrained by reproducibility and alignment challenges. The main translational bottleneck is therefore not whether saliva changes with exercise stress, but whether the workflow can separate biology from pre-analytical and analytical drift. The most credible route toward decision-grade athlete monitoring is a platform-aware workflow that combines standardized passive-drool sampling, explicit preprocessing and extraction logic, repeated within-athlete baselines, targeted or pseudo-targeted validation panels, and multimodal integration with performance and training data.