Machine learning approaches in predicting circadian rhythm sleep-wake disorders: A review

Area of Science:

• Chronobiology
• Computational Biology
• Sleep Medicine

Background:

• Circadian rhythm sleep-wake disorders (CRSWDs) disrupt sleep health due to misalignment between internal body clocks and external schedules.
• Machine learning (ML) offers advanced analytical capabilities for complex biological data, yet its use in CRSWD prediction is nascent.
• Digital phenotyping and sleep-omics generate rich datasets relevant to circadian rhythm research.

Purpose of the Study:

• To systematically review and synthesize existing research on ML applications for human circadian phase estimation.
• To evaluate the efficacy of ML algorithms in classifying CRSWDs using diverse data sources.
• To identify limitations and future directions for ML in circadian medicine.

Main Methods:

• Systematic literature search across major databases (PubMed, ScienceDirect, PsycINFO, Cochrane, Google Scholar) from January 2011 to May 2025.
• Inclusion criteria focused on ML models predicting circadian phase or classifying CRSWDs in sighted adults.
• Data sources included sleep diaries, actigraphy, genomics, and biological markers; pediatric, neurodegenerative, and blind populations were excluded.

Main Results:

• Twenty-two studies met the inclusion criteria, demonstrating ML's potential in circadian phase estimation and CRSWD classification.
• ML models successfully utilized various data modalities, including wearable sensor data and multimodal features.
• Significant limitations were noted, including small sample sizes, methodological inconsistencies, and insufficient external validation.

Conclusions:

• ML-based approaches show considerable potential for improving CRSWD detection and enabling personalized circadian medicine.
• Current evidence highlights the need for more rigorous, standardized, and generalizable ML models.
• Future research should focus on large, multimodal datasets and transparent validation for scalable clinical tools.