Analysis of wearable time series data in endocrine and metabolic research

Figure 1. Endocrine and metabolic rhythms have been widely observed, and are key to sustain healthy states. Left: Misalignment of rhythms in signalling molecules (e.g., melatonin, cortisol, and glucose) is strongly associated to physio-pathological changes such as dietary changes, pregnancy, shift work, social jet lag, systemic inflammation and chronic disease. Top right: Wearable device signals (e.g., heart rate (HR), body temperature, and accelerometer-derived actigraphy) may act as continuous surrogates of endocrine rhythms specific to an individual's internal time (e.g., chronotype, glucotype). Bottom right: Endocrine and wearable signals are commonly described using metrics such as dim light melatonin onset (DLMO), cortisol awakening response (CAR), heart rate variability (HRV) and area under the curve (AUC). Mathematical and computational techniques, including machine learning, can help identify these metrics as clinical proxies (e.g., computational biomarkers) to support early diagnosis and management of disease.