ABSTRACT
Biological systems usually consist of large number of components and involve processes at a variety of spatial, temporal and biological scales. This study presents a framework for model identification and the use of Global Sensitivity Analysis (GSA) in systems biology modelling and shows how the information content of clinical data from Short Insulin Tolerance Test (SITT) can be handled by optimal model-based estimation techniques. The goal is to identify dynamic model of type II diabetes and estimate a set of parameters of the model with greater accuracy and precision. Based on the SITT data, the blood glucose dynamic model was identified as a system of linear differential equation with constant coefficients (parameters). The sensitivity of the parameters was tested using a novel GSA based approach, and Derivative-Based Global Sensitivity Measures (DGSM). The proposed approach was implemented in SensSB (a Matlab based toolbox). For the purpose of comparison, the sensitivity of the model was also tested using Sobol's method and a local approach. The results have shown that the model is less sensitive to the third parameter ( ) and the model fits the SITT data satisfactorily. Subsequently, a control strategy called receding horizon control was investigated to regulate blood glucose concentration under the model predictive control framework. Two forms of receding horizon control strategy (fixedend and moving-end) were proposed and applied to the dynamic model to maintain blood glucose concentration. Different disturbance scenarios were generated to evaluate the performance of the two strategies in terms of its efficiency to handle disturbances. The control strategies successfully addressed the issues of the input/external disturbance considered for the patients in a virtual situations which maintain blood glucose level at 80.06 mg/dL. 2 K