Download files
Abstract
There has been a dramatic global increase in the incidence of type 2 diabetes.
As type 2 diabetes results from dysregulation of the control mechanism of
glucose homeostasis, it is not amenable to purely reductionist investigative
methods. The methods of control theory and mathematical modelling, which
are uniquely suited to obtaining system-wide insights, are increasingly being
used to investigate complex biological questions such as the etiology of
diabetes.
The Sedaghat, Sherman and Quon model is a widely-cited mathematical
model of insulin signalling published in 2002. The model captures many
important signalling mechanisms and their interactions, however, it is also
known to have some limitations. In this study, a local parametric sensitivity
analysis (PSA) based on the time integral of GLUT4 expression at the plasma
membrane (a measure related to glucose transport) was used to investigate
the Sedaghat model. Sensitivity profiles for all rate constants and initial
conditions over a range of insulin concentrations, input profiles and parameter
perturbations are presented.
The PSA revealed several important features of the model. There was an obvious
saturation phenomenon at high insulin levels that affcted much of the
network. The sensitivity of many parameters changed substantially across
the insulin concentration range. These two features highlight the fact that results
obtained under high insulin conditions, either in vivo or in silico, cannot
necessarily be extrapolated to the physiological range. Furthermore, parameters
from the model were classiffied as either sensitive (having a substantial
influence on the model output) or insensitive (having little discernible effect).
The major locations of regulation in the signalling network were determined
by these means and flaws in the network identified. Potential improvements
to the model were also highlighted by the PSA.
In view of the model's limitations, and the biological knowledge gained over
the past ten years, it is clear that the model as a whole is in need of major
structural improvements. This work shows that we now have the analytical
tools to attempt this task.