A night without sleep leaves certain chemical traces in saliva that can be reliably detected with high accuracy. A recent study published in the Journal of Proteome Research provides evidence that acute sleep deprivation creates a unique pattern of molecules in the mouth, ultimately paving the way for rapid tests that can identify exhausted drivers and workers. This study suggests that as few as 10 to 12 specific biomarkers are sufficient to distinguish between those who are awake 24 hours a day and those who are well rested.
A team of scientists from the University of Zurich in Switzerland designed this clinical trial to find biological markers of sleep deprivation. A research group led by authors Michael Scholz and Thomas Kraemer aimed to characterize the oral fluid metabolome in different states of fatigue.
Sleep deprivation is a growing problem that negatively impacts public health, productivity, and road safety. Currently, police officers and employers must rely on self-reported sleep habits and subjective observations to determine whether someone is too tired to operate a vehicle or heavy equipment. Unlike alcohol intoxication, which can be easily measured with a standard breathalyser, there is no simple, direct, objective way to measure sleep deprivation.
To address this gap, the authors turned to a field called metabolomics. Metabolomics is the study of small molecules, known as metabolites, left behind in biological samples as functional end products of cellular processes. The concentrations of these chemicals change each time your body uses energy, repairs tissue, or responds to stress.
The researchers suspected that extreme fatigue might alter the body’s chemistry, creating a recognizable metabolic footprint. They specifically chose to test oral fluid, or saliva. Because it is non-invasive and easy to harvest. Blood sampling can be inconvenient or legally complex for non-medical personnel to perform on the street, making saliva an ideal vehicle for future point-of-care testing.
“Our study provides for the first time a direct biomarker of sleep deprivation in saliva under realistic conditions. This is a milestone for forensic research,” said Thomas Kremer, professor of forensic pharmacology and toxicology at the UZH Institute of Forensic Medicine.
To test their hypothesis, the scientists recruited 20 healthy young men with an average age of about 24 years. The study focused specifically on young, normal-weight men because demographic data show that they are the group most at risk for road accidents due to sleepiness. All participants reported a habitual sleep duration of 7 to 9 hours each night, with no extreme morning or evening preferences.
This trial used a randomized crossover design. That is, each participant had at least one week of normal sleep between each session and experienced three different sleep states in random order. This condition was designed to replicate common real-world sleep scenarios. In the control condition, participants were allowed a standard 8 hours of sleep.
In the two experimental conditions, participants accumulated the same 8-hour sleep deficit, but achieved this deficit in two different ways. Completely sleep-deprived, the men had no sleep all night and had to stay awake for more than 24 hours straight. In the sleep restriction condition, the men had to reduce their normal sleep time by two hours each night for four consecutive nights, resulting in six hours of sleep each night.
During each intervention, researchers collected unstimulated saliva samples at predefined times throughout the day and evening. Participants placed a small cotton swab under their tongue for 2 minutes without chewing to ensure the collection was consistent and free of contamination. To control for natural biological rhythms, the scientists also measured the participants’ melatonin expression in dim light. Tracking the release of this natural hormone provided a reliable way to map each individual’s biological clock.
The collected samples were analyzed using a technique called liquid chromatography combined with high-resolution mass spectrometry. This advanced analytical method separates the complex mixture of chemicals found in saliva and identifies them based on their mass and charge. This process resulted in a large dataset containing over 6,000 robust molecular features.
To make sense of this vast amount of data, the authors employed interpretable machine learning techniques. They trained a logistic regression model, a mathematical algorithm used to predict outcomes, to classify whether unidentified saliva samples belonged to a resting state, sleep deprivation state, or sleep restriction state. The algorithm is programmed to recognize patterns in metabolic data without requiring a baseline sample from the same individual for comparison. This reference-free approach is essential for real-world forensic applications where baseline samples are rarely available.
“We found that acute sleep deprivation affects approximately 10% of all biomolecules in saliva. The challenge was to identify among tens of thousands of molecules that reliably signal fatigue. Using cutting-edge technology, we were able to identify 10 biomarkers that do just that,” said lead author Michael Scholz.
The analysis revealed that the sudden loss of a night’s sleep produces a unique and very distinct metabolic fingerprint. The machine learning model was able to detect complete sleep deprivation with remarkable accuracy using a reduced set of just 10 to 12 molecular features. If the model flagged a sample as coming from a sleep-deprived donor, its prediction was correct about 96% of the time.
Interestingly, the predictive power of the molecular fingerprints tended to vary depending on the time of day. The metabolic differences between rested and exhausted participants were most pronounced during the morning and midday hours. By late evening, the chemical signatures began to converge. This is likely because the natural circadian rhythms that promote nighttime sleepiness temporarily masked certain markers of prolonged wakefulness.
Complete sleep deprivation left a clear chemical imprint, but four nights of sleep restriction did not cause any exploitable metabolic changes. The algorithm struggled to reliably distinguish between participants who slept six hours a night and those who slept eight hours a night. This suggests that the body may process chronic, moderate sleep deprivation differently than acute, sudden, sustained wakefulness. The authors suspect that more sleep debt needs to accumulate to trigger the specific metabolic fingerprint associated with extreme fatigue.
The ability to detect 24-hour alertness has important practical applications for law enforcement. In some areas, such as New Jersey, certain laws classify driving while awake for 24 consecutive hours as reckless driving. The model developed in this study demonstrates the theoretical possibility of enforcing such legal standards by providing objective evidence of severe attrition.
“Tests like this have the potential to improve road safety and make work environments safer where alertness and concentration are important,” Scholz said.
Despite these promising advances, the authors note that there are several limitations that require further investigation. The current study investigated only healthy young men who adhered to a regular day and night schedule. Future studies should test these molecular markers in women, older adults, and individuals with different chronotypes and sleep disorders so that the results can be generalized to broader populations.
Additionally, biomarkers need to be validated under a broader range of real-world scenarios. Scientists need to investigate how factors such as shift work, dietary changes, alcohol intake, and prescription medications can alter the oral metabolome and interfere with test accuracy. Determining the precise chemical structures and names of 10 to 12 essential metabolites is also a necessary next step.
It is also important to recognize the biological difference between sleep deprivation and functional impairment. Current machine learning models are designed to detect the physiological state of being awake 24 hours a day, but this does not automatically and perfectly correlate with an individual’s cognitive impairment. Tolerance to sleep deprivation varies from person to person. Some people exhibit chemical markers of fatigue while maintaining adequate performance, while others develop severe impairments after just a few hours of sleep deprivation.
Ultimately, this proof-of-concept study demonstrated that bibliometric-free, metabolomics-based sleep deprivation detection is theoretically possible. As research progresses, these findings will lay the foundation for reliable, non-invasive tools that may eventually be deployed by police officers and workplace safety managers.
The study, “Forensic Exploitation of the Metabolic Fingerprint of Sleep Deprivation and Sleep Restriction: A Machine Learning Study in Oral Fluid Metabolomics,” was authored by Michael Scholz, Andrea E. Steuer, Akos Dobay, Hans-Peter Landolt, and Thomas Kraemer.
