Computational Modelling of Glucocorticoid Chronotherapy for Diseases with Cortisol Circadian Dysregulation

Abstract Unnatural exposure to nighttime light, shift work, and chronic disturbance in sleep/ wake and feeding cycles are known to cause circadian rhythm (CR) disruption of the hypothalamic-pituitary axis. Dysregulated cortisol CR is associated with metabolic dysfunction, sleep disorders, mood, and cognitive dysfunction. Addressing cortisol circadian dysregulation to restore normal peripheral physiology could be a potential option. We used an ODE-based computational model to study the complexity of the HPA-Inflammation system and the time-dependent effects of glucocorticoids on systems CR. Further, the model was validated with the reported human circadian physiology of system state variables. A treatment response analysis (TRA) based on phase and amplitude responses of the state variable was used to identify chronotherapeutic dosing of dexamethasone (Dex) to correct cortisol CR. The TRA suggest optimal time ranges for Dex dosing: 5–6 AM for Hypocortisolemia and 8–9 AM for lower-grade inflammation. Several cortisol rhythm defects are simulated to generate training data for a neural network (NN) to predict the optimal time for Dex dosing in cases of cortisol CR dysregulation. The NN protocol forecasted chronotherapeutic dosing for psychiatric, neuroendocrine, and inflammatory diseases involving cortisol CR dysfunction, such as hyper- or hypocortisolemia and CR phase shifts. We demonstrate a hybrid method in which mechanistic models generate detailed synthetic clinical data to train NN models to predict glucocorticoid chronotherapy, a task that would otherwise be difficult to perform clinically.