0009 Circadian Proteins and Circadian Phase Estimation in Healthy Adults Free

Circadian rhythms, which control sleep-wake cycles and metabolism, are fundamental to human health. Our study aimed to understand how these rhythms affect proteins in the body throughout the day, and to develop a tool that predicts the body's internal clock phase based on protein expression.

Plasma samples from 17 healthy adults were collected hourly under controlled conditions designed to unmask endogenous circadian rhythmicity; in a subset of 8 participants, we also collected samples across a day on a typical sleep-wake schedule. Using the SomaScan aptamer-based multiplexed platform, we analyzed a total of 6916 proteins were analyzed. We used differential rhythmicity analysis based on a cosinor model with mixed effects to identify a subset of proteins that demonstrates circadian rhythmicity. Finally, we trained a machine learning model to predict the Dim Light Melatonin Onset (DLMO) for a given protein sample.

Four hundred and thirty-one (6.2%) proteins displayed consistent endogenous circadian rhythms on both a sleep-wake schedule and under controlled conditions. This subset not only aligns with the proteins selected by the elastic-net model but also maintains performance without diminishing it in comparison to using the complete set of available proteins. Overall, our circadian phase predictor reached a median absolute error (MdAE) of 1.2 hours when performing a leave-one-cross-out cross-validation subject-wise.

This research demonstrates that a considerable number of plasma proteins follow natural circadian rhythms within the human body. Furthermore, it establishes that the DLMO can be accurately predicted with a MdAE slightly over one hour using a single blood sample. More research with larger and more diverse datasets is essential to confirm our method, promising better treatments for circadian disorders and advancing personalized circadian health care.

Collection, processing, and analysis of the samples were supported by National Institutes of Health (NIH) grant R01 HL148704 and unrestricted gifts. Collection of additional samples was supported by US Office of Naval Research grant N00014-15-1-2408. The BWH CCI is supported by Harvard Catalyst | The Harvard Clinical and Translational Science Center (National Center for Advancing Translational Sciences, NIH Award UL1 TR002541) and financial contributions from Harvard University and its affiliated academic healthcare centers.