Microphysiological systems (MPS) are powerful tools for emulating human physiology and replicating
disease progression in vitro. MPS could be better predictors of human outcome than current animal models, but mechanistic interpretation and in vivo extrapolation of the experimental results remain significant challenges. Here, we address these challenges using an integrated experimental-computational approach. This approach allows for in silico representation and predictions of
glucose metabolism in a previously reported MPS with two organ compartments (liver and pancreas) connected in a closed loop with circulating medium. We developed a computational model describing
glucose metabolism over 15 days of culture in the MPS. The model was calibrated on an experiment-specific basis using data from seven experiments, where HepaRG single-liver or liver-islet cultures were exposed to both normal and hyperglycemic conditions resembling high
blood glucose levels in diabetes. The calibrated models reproduced the fast (i.e. hourly) variations in
glucose and
insulin observed in the MPS experiments, as well as the long-term (i.e. over weeks) decline in both
glucose tolerance and insulin secretion. We also investigated the behaviour of the system under
hypoglycemia by simulating this condition in silico, and the model could correctly predict the
glucose and
insulin responses measured in new MPS experiments. Last, we used the computational model to translate the experimental results to humans, showing good agreement with published data of the
glucose response to a meal in healthy subjects. The integrated experimental-computational framework opens new avenues for future investigations toward disease mechanisms and the development of new
therapies for metabolic disorders.