Somnotate: A probabilistic sleep stage classifier for studying vigilance state transitions
Paul J. N. Brodersen,Hannah Alfonsa,Lukas B. Krone,Cristina Blanco-Duque,Angus S. Fisk,Sarah J. Flaherty,Mathilde C. C. Guillaumin,Yi-Ge Huang,Martin C. Kahn,Laura E. McKillop,Linus Milinski,Lewis Taylor,Christopher W. Thomas,Tomoko Yamagata,Russell G. Foster,Vladyslav V. Vyazovskiy,Colin J. Akerman
DOI: https://doi.org/10.1371/journal.pcbi.1011793
2024-01-18
PLoS Computational Biology
Abstract:Electrophysiological recordings from freely behaving animals are a widespread and powerful mode of investigation in sleep research. These recordings generate large amounts of data that require sleep stage annotation (polysomnography), in which the data is parcellated according to three vigilance states: awake, rapid eye movement (REM) sleep, and non-REM (NREM) sleep. Manual and current computational annotation methods ignore intermediate states because the classification features become ambiguous, even though intermediate states contain important information regarding vigilance state dynamics. To address this problem, we have developed "Somnotate"—a probabilistic classifier based on a combination of linear discriminant analysis (LDA) with a hidden Markov model (HMM). First we demonstrate that Somnotate sets new standards in polysomnography, exhibiting annotation accuracies that exceed human experts on mouse electrophysiological data, remarkable robustness to errors in the training data, compatibility with different recording configurations, and an ability to maintain high accuracy during experimental interventions. However, the key feature of Somnotate is that it quantifies and reports the certainty of its annotations. We leverage this feature to reveal that many intermediate vigilance states cluster around state transitions, whereas others correspond to failed attempts to transition. This enables us to show for the first time that the success rates of different types of transition are differentially affected by experimental manipulations and can explain previously observed sleep patterns. Somnotate is open-source and has the potential to both facilitate the study of sleep stage transitions and offer new insights into the mechanisms underlying sleep-wake dynamics. Typically, the three different vigilance states–awake, REM sleep, and non-REM sleep–exhibit distinct features that are readily recognised in electrophysiological recordings. However, particularly around vigilance state transitions, epochs often exhibit features from more than one state. These intermediate vigilance states pose challenges for existing manual and automated classification methods, and are hence often ignored. Here, we present 'Somnotate'—an open-source, highly accurate and robust sleep stage classifier, which supports research into intermediate states and sleep stage dynamics in mice. Somnotate quantifies and reports the certainty of its annotations, enabling the experimenter to identify abnormal epochs in a principled manner. We use this feature to identify intermediate states and to detect unsuccessful attempts to switch between vigilance states. This has the potential to provide new insights into the mechanisms of vigilance state transitions, and creates new opportunities for future experiments.
biochemical research methods,mathematical & computational biology