A DIGITAL TWIN FRAMEWORK FOR EXERCISE-NUTRITION INTERACTION MODELING: AN IMMUNOMETABOLIC PERSPECTIVE

This work investigates the physiological basis of post-prandial glucose regulation by considering gastrointestinal dynamics, hormonal control, and exercise, with a specific focus on immunometabolic signaling mediated by interleukin-6 (IL-6) and glucagon-like peptide-1 (GLP-1). The interaction between exercise and nutrition represents a physiologically relevant yet insufficiently modeled scenario, in which the timing of exercise relative to meal ingestion can substantially influence post-prandial glucose dynamics. The aim is to develop a mechanistic digital twin capable of quantifying the effects of pre-prandial exercise on post-prandial glucose dynamics. The proposed approach combines exercise-driven IL-6/GLP-1 signaling, macronutrient-specific gastric emptying and absorption, and glucose-insulin regulation into a system of ten ordinary differential equations (ODEs), enabling a representation of how exercise modulates digestive timing and glucose appearance in plasma. Model personalization is achieved using data from CGMacros PhysioNet dataset, after selection of metabolically healthy subjects (n = 6). Subject-specific parameters are identified from standardized breakfast and applied to predict responses to meals with different nutritional compositions, demonstrating generalization capabilities. The framework is then employed to isolate the effects of pre-prandial exercise through controlled ON/OFF simulations, revealing consistent exercise-induced delays in gastric emptying and glucose rate of appearance across individuals. These temporal shifts are summarized using the time-to-50% area metric (t50), corresponding to mean delays of approximately ≈ 20.75 ± 12.16 min for gastric emptying and ≈ 17.64 ± 11.43 min for glucose appearance. Overall, the results support the interpretation of exercise as a modulator of post-prandial glucose dynamics primarily acting through delayed gastrointestinal transit. The proposed framework is positioned as an offline, mechanistic digital twin for replay and what-if analyses, providing an interpretable foundation for future extensions toward richer biomarker integration, improved identifiability of macronutrient mechanisms, and adaptive approaches aimed at fully personalized and clinically relevant digital twins.