ELS rats display REM-like oscillatory dynamics during active behavior.
A, State map construction reflects the sleep-wake cycle and spectral dynamics for one representative CTRL (top) and ELS (bottom) rat. PFC delta power is evident during NREM, theta/delta ratio is maximum in REM, EMG power describes ACT states, and HPC-PFC theta coherence is stronger close to REM. Note that in the ELS rat, the coherence is also broadly high during the ACT state. B, Representative state maps from three CTRL (top) and three ELS (bottom) rats. CTRL rats displayed a clear separation between REM (red), ACT (light green), and NREM (blue) clusters. ELS displayed an overlap between REM and ACT states. C, Distributions of the Euclidean distances between each REM and ACT epoch for CTRL (top) and ELS (bottom) rats. ELS distributions were heterogeneous, many of which were skewed to shorter distances. D, Average Euclidean distance showing that ELS rats have REM-like oscillatory dynamics during active behavior. E, REM-like oscillations during ACT were more pronounced in ELS during the initial 12 h of recording. F, Machine learning approach to decode REM versus ACT states. We used radial basis function support vector machine (SVM) and random forest (RF) algorithm in order to classify REM and ACT epochs in CTRL (black) or ELS (orange) rats. Using a bootstrapped confidence interval, we observed that the discriminative performance was worse in ELS animals, even when using “brute-force” algorithms. G, A machine learning algorithm (random forest) fitted on whole1572 spectrum data also shows a significantly worse performance in ELS for PFC power, HPC power, coherence, and all estimates combined. † p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001. CTRL n=6, and ELS n=9. Error bars represent the mean ± SEM.