Fig. 7. Dynamic modeling and analyses.

a Scheme of parameter estimation. After data acquisition, parameters are fitted in models by minimizing the difference between experimental data and model output. Sensitivity analysis, uncertainty quantification and identifiability analysis help assess the performance and robustness of the fit. b Loop of Model Predictive Control. Based on the output of model prediction, this control strategy updates control input (e.g., drug administration) to make the system dynamics track reference trajectory (e.g., desired tumor cell decrement) during each time interval. Its essence is to handle the constrained optimization problem (Constraints can be maximal drug doses and minimal normal cell populations).