Archetypal Analysis for the Characterization of Distinct Patterns in Mouse Sleep Data

Area of Science:

• Neuroscience
• Computational Biology
• Data Science

Background:

• Current sleep staging may oversimplify complex sleep-wake dynamics.
• Sleep is increasingly understood as a continuum rather than discrete states.

Purpose of the Study:

• To explore archetypal analysis (AA) for extracting distinct sleep characteristics.
• To represent sleep data as a continuum using AA.
• To assess AA's ability to capture sleep stages, mouse type, and individual variability.

Main Methods:

• Applied AA to EEG power and EMG RMS data from wild-type and narcolepsy mice.
• Analyzed sample-specific profiles and their continuum of archetypes.
• Quantified correspondences using normalized mutual information (NMI) and permutation testing.

Main Results:

• AA robustly extracted distinct sleep data characteristics.
• AA-derived profiles showed correspondence to sleep stages, mouse type, and ID.
• Correspondences were significant but low (NMI≤0.2), indicating richer data properties.

Conclusions:

• AA provides interpretable characterizations of sleep data, representing samples as a continuum.
• This approach offers a versatile tool for understanding sleep structure and properties in health and disease.
• AA enables novel, interpretable characterization of sleep data, enhancing understanding of sleep in various conditions.