Leveraging the Metabolic Fingerprint of Sleep Deprivation and Sleep Restriction for Forensic Applications: A Machine...

As sleep loss leads to accidents and impaired safety, a direct metabolic marker would be beneficial for forensic interpretation. In a sufficiently powered, randomized, controlled, crossover trial under realistic conditions, we examined the salivary metabolome of 20 young men (habitual sleep duration 7-9 h) following three interventions: one night of total sleep deprivation, four consecutive nights of sleep restriction to 6 h, and control (8 h of sleep). Oral fluid specimens were repeatedly collected and analyzed using liquid chromatography coupled to mass spectrometry. Logistic regression models were trained to classify unseen samples without reference samples from the same individual. Acute sleep deprivation exhibited a unique metabolic fingerprint that could be detected precisely (F_0.5 = 0.90) when using only 12 molecular features. This fingerprint was more pronounced in samples collected in the morning/midday hours. Nevertheless, at all time points, the overall correct predictions by far outweighed the incorrect ones. Four nights of sleep restriction did not lead to exploitable metabolic changes. This study presents a metabolic fingerprint of acute sleep deprivation in oral fluid under realistic conditions and explores practical implications and limitations of its machine learning-aided classification. Metabolomics-based, reference-free sleep loss detection holds potential for applications in forensic, clinical, and occupational contexts.