A machine-learning algorithm integrating baseline serum proteomic signatures predicts exercise responsiveness in overweight males with prediabetes
Candela Díaz-Cañestro, Jiarui Chen, Yan Liu, Hao Han, Yao Wang, Éric Honoré, Chi‐Ho Lee, Karen S.L. Lam, Michael Tse, Aimin Xu
Abstract
The molecular transducers conferring the benefits of chronic exercise in diabetes prevention remain to be comprehensively investigated. Herein, serum proteomic profiling of 688 inflammatory and metabolic biomarkers in 36 medication-naive overweight and obese men with prediabetes reveals hundreds of exercise-responsive proteins modulated by 12-week high-intensity interval exercise training, including regulators of metabolism, cardiovascular system, inflammation, and apoptosis. Strong associations are found between proteins involved in gastro-intestinal mucosal immunity and metabolic outcomes. Exercise-induced changes in trefoil factor 2 (TFF2) are associated with changes in insulin resistance and fasting insulin, whereas baseline levels of the pancreatic secretory granule membrane major glycoprotein GP2 are related to changes in fasting glucose and glucose tolerance. A hybrid set of 23 proteins including TFF2 are differentially altered in exercise responders and non-responders. Furthermore, a machine-learning algorithm integrating baseline proteomic signatures accurately predicts individualized metabolic responsiveness to exercise training.