Unravelling the insulin signalling pathway using mechanistic modelling

Type two diabetes affects 5% of the world's population and is increasing in prevalence. A key precursor to this disease is insulin resistance, which is characterised by a loss of responsiveness to insulin in liver, muscle and adipose tissue. This thesis focuses on understanding insulin signalling using the 3T3-L1 adipocyte cell model. Computational modelling was used to generate quantitative predictions in the signalling pathways of the adipocyte, many of which are mediated by enzymatic reactions. This study began by comparing existing enzyme kinetic models and evaluating their applicability to insulin signalling in particular. From this understanding, we developed an improved enzyme kinetic model, the differential quasi-steady state model (dQSSA), that avoids the reactant stationary assumption used in the Michaelis Menten model. The dQSSA was found to more accurately model the behaviours of enzymes in large in silico systems, and in various coenzyme inhibited and non-inhibited reactions in vitro. To apply the dQSSA, the SigMat software package was developed in the MATLAB environment to construct mathematical models from qualitative descriptions of networks. After the robustness of the package was verified, it was used to construct a basic model of the insulin signalling pathway. This model was trained against experimental temporal data at 1 nM and 100 nM doses of insulin. It revealed that the simple description of Akt activation, which displays an overshoot behaviour, was insufficient to describe the kinetics of substrate phosphorylation, which does not display the overshoot behaviour. The model was expanded to include Akt translocation and the individual phosphorylation at the 308 and 473 residues. This model resolved the discrepancy and predicts that Akt substrates are only accessible to Akt localised in the cytosol and that PIP3 sequestration of cytosolic Akt acts as a negative feedback