Abstract
Determining physiological stress at high resolution is crucial across diverse settings to enable informed decision-making in the context of health and disease. Saliva-based targeted multi-omics testing provides a powerful, non-invasive method to quantify physiological stress and circadian dynamics at high-frequency. In a laboratory crossover trial with 24-hour sampling comprising 413 saliva samples, we demonstrate high analytical reliability, distinct molecular individuality, and robust acute and delayed responses to physical exercise across proteins, metabolites, and lipids. Moreover, we present the most comprehensive existing dataset describing 24-hour molecular kinetics across these three omics layers. Leveraging this controlled setting, we applied machine learning to single-timepoint saliva samples to accurately predict recent physical exercise both immediately after and 24 hours later. Next, we translated this analytical framework to a real-world longitudinal setting of elite football players monitored over 16 months, comprising over 12,000 saliva samples. Despite increased biological and contextual variability, the model retained robust discrimination between exercise and rest on the following day. Based on prediction probabilities, we introduce a saliva-based internal strain metric, that captures internal load and can be harnessed to monitor physical exercise and recovery. Model robustness was further supported through out-of-sample validation using previously unseen observations. Our findings demonstrate that saliva-based targeted multi-omics reliably captures physical exercise and recovery states in both laboratory and real-world environments, providing a scalable framework for monitoring physical performance. This non-invasive approach holds broad potential for physiological monitoring and can serve as a blueprint for health- and disease-related contexts.
Introduction
Quantifying how the human body responds to and recovers from stress—whether induced by elite athletic performance, disease, or daily life—is a cornerstone of personalized medicine. Traditionally, such monitoring has required invasive blood draws, which are difficult to perform at the high frequencies needed to capture dynamic physiological shifts. While wearable technologies provide valuable "external" load data (e.g., GPS, heart rate), they often fail to reflect the "internal" integrated biological response across multiple systems.
Saliva has emerged as a promising, non-invasive alternative for frequent sampling. However, a comprehensive understanding of how the salivary molecular landscape changes over a 24-hour cycle and in response to intense physical exertion has been lacking. This study addresses this gap by establishing a targeted multi-omics framework to map physiological stress and recovery.
Methods
The researchers utilized a targeted multi-omics approach, measuring specific sets of proteins, metabolites, and lipids using Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS). The study integrated two distinct cohorts to validate the methodology across controlled and real-world environments.
The Laboratory Cohort involved 12 healthy female participants in a randomized crossover trial, generating 413 samples with bi-hourly sampling over 24 hours to establish baseline circadian rhythms and acute exercise responses. The Real-World Cohort followed 49 elite male football players monitored over 16 months, resulting in a dataset of over 12,000 samples.
To analyze this high-dimensional data, a machine learning pipeline was developed. Dimensionality reduction (t-SNE) visualized individual "molecular fingerprints," while Bayesian Additive Models (BAMs) mapped 24-hour natural fluctuations. Classification models (Random Forest, XGBoost, etc.) were trained to distinguish "exercise" from "rest" states, and SHAP (SHapley Additive exPlanations) values identified key biomarkers.
Key Insights and Results
Molecular Individuality and Circadian Rhythms
A significant finding was the high degree of molecular individuality in saliva. Even following intense exercise, an individual's basic molecular signature remained stable enough for identification (Figure 1b). Additionally, the study provided the most comprehensive dataset to date of 24-hour saliva kinetics, revealing clear circadian rhythms in many analytes previously only tracked in blood, such as certain bile acids and lipids.
Predictive Performance and the "24-Hour Echo"
The machine learning models demonstrated high accuracy in detecting physiological stress. In the laboratory setting, the model achieved an ROC-AUC of 0.95 for immediate exercise detection and 0.85 for detection 24 hours later. This "24-hour echo" suggests that the biological impact of exercise persists in saliva long after traditional markers like heart rate have normalized. In the "messy" real-world environment of professional sports, the model maintained a robust ROC-AUC of 0.84 for next-day exercise prediction.
Internal Strain Metric (ISM)
The researchers introduced the Internal Strain Metric (ISM), a continuous score (0 to 1) derived from ML prediction probabilities. A high ISM indicates a "strained" state, while a low score suggests recovery. The ISM showed a statistically significant decrease as recovery time increased (p = 0.000105), providing a quantitative tool for tracking the physiological "winding down" process.
Discussion
This study proves that saliva-based targeted multi-omics can reliably capture complex physiological states previously thought to be visible only in blood. The identification of 13 key exercise biomarkers, including Cystatin-C and 2-trans-octenoylcarnitine, provides a foundation for non-invasive performance monitoring.
While promising, the study notes limitations, such as the divergence between the laboratory (female) and real-world (male) cohorts. Future research should validate these models across broader demographics and compare the ISM against established strain measures like heart rate variability (HRV). Ultimately, transitioning these laboratory-based techniques to portable, rapid-testing platforms could revolutionize physiological monitoring in sports, clinics, and daily life.
Summary updated