The 103,200-arm acceleration dataset in the UK Biobank revealed a landscape of human sleep phenotypes

Abstract

Human sleep phenotypes can be defined and diversified by both genetic and environmental factors. However, some sleep phenotypes are difficult to evaluate without long-term, precise sleep monitoring, for which simple yet accurate sleep measurement is required. To solve this problem, we recently developed a state-of-the-art sleep/wake classification algorithm based on wristband-type accelerometers, termed ACCEL (acceleration-based classification and estimation of long-term sleep-wake cycles). In this study, we optimized and applied ACCEL to large-scale analysis of human sleep phenotypes. The clustering of an about 100,000-arm acceleration dataset in the UK Biobank using uniform manifold approximation and projection (UMAP) dimension reduction and density-based spatial clustering of applications with noise (DBSCAN) clustering methods identified 16 sleep phenotypes, including those related to social jet lag, chronotypes (“morning/night person”), and seven different insomnia-like phenotypes. Considering the complex relationship between sleep disorders and other psychiatric disorders, these unbiased and comprehensive analyses of sleep phenotypes in humans will not only contribute to the advancement of biomedical research on genetic and environmental factors underlying human sleep patterns but also, allow for the development of better digital biomarkers for psychiatric disorders.