Multi-Methodological Characterization of Sleep Deprivation: From Standard EEG Power Spectra to Aperiodic Dynamics in Humans and Mice

Sleep deprivation is a potent, rapid-acting therapeutic intervention for major depressive disorder, yet its underlying neural mechanisms remain poorly understood, hindering the development of predictive biomarkers. Here, we systematically characterize the electrophysiological signatures of prolonged wakefulness using a multi-methodological approach across three independent datasets in humans and mice. By integrating standard power spectral density analysis with aperiodic component fitting (SpecParam) and highly comparative time-series analysis (HCTSA), we identified robust cross-species biomarkers of sleep pressure. Machine learning models revealed that theta power is the most consistent feature for differentiating control and sleep deprivation states, achieving up to 90% classification accuracy. Sleep deprivation significantly increased the spectral offset - suggesting global cortical hyperexcitation - while simultaneously steepening the spectral slope. We interpret this simultaneous shift as a state uncoordinated state of hyperexcited and inefficient neural processing. These findings establish reproducible EEG markers of sleep deprivation that transcend species. Given the clinical utility of wake therapy, we propose that prefrontal theta power and spectral offset/slope may serve as mechanism-based predictors of therapeutic response. Our results provide a framework for the clinical validation of these biomarkers, potentially enabling personalized chronotherapeutic interventions for psychiatric disorders.