Fig. 3

Results of model training and validation on Fast-food study data. A. Comparison between model uncertainty (light purple area) for the best model simulation (the dark purple line) with the training data (purple error bars) or validation data (grey error bars). On the whole-body level, data for B weight and C fat mass (\(\mathrm{FM}\)) and fat free mass (\(\mathrm{FFM}\)) was used for training and validation. On the tissue/organ level, data for D glucose and insulin was used for training and validation. E The diabetes effects on the cell level model–decrease in \(\mathrm{IR}\), decrease in \(\mathrm{GLUT}4\), and \(\mathrm{diabetes}\), representing an attenuation of. F Scaling of the three diabetes parameters (with the chosen values indicated with triangles) and the resulting behavior of the simulation curves as dose responses to insulin, to match the fit to data in G. Data and simulations of the dose responses of phosphorylated \(\mathrm{PKB}\), \(\mathrm{PKB}308-\mathrm{p}\) and phosphorylated \(\mathrm{IRS}1\), \(\mathrm{IRS}1-\mathrm{p}\) in response to the indicated concentrations of insulin for 10 min and normalized 0–100%. The predicted simulation before the Fast-food diet (blue solid line) use the non-diabetic parameters from [7] as they were, which gave a good agreement with data (blue error bars with circles). The three diabetes parameters were scaled to get the predicted simulation after the diet (purple dashed line) to fit to the corresponding data (purple error bars with squares)