1.2.2 Sensitivity Analysis
Sensitivity analysis is a technique that determines how changes in the
values of parameters affect a given response variable of interest.
Originally, sensitivity analysis was employed to determine the
robustness of a system to variability or uncertainty in parameter
values. This has been extended to guide model-order reduction by
determining the system’s state that has the least influence on system
outputs. In addition to reducing the number of nodes (compartments),
sensitivity analysis can also be used as an edge (reactions) reduction
method to determine which reactions contribute least to the output of
the system and eliminate those reactions by setting their associated
rate parameters to zero [44].
Several algorithms for global sensitivity analysis have been developed
with Fourier amplitude and Sobol’s sensitivity analysis being common
[45,46]. Jayachandran et al. employed Sobol’s method and achieved a
50% reduction in model parameters for a model of 6-mercaptopurine
effects in children with acute lymphoblastic leukaemia. The technique
enabled the estimation of model parameters for individual patients and
subsequent use in treatment individualisation [21]. Gueorguieva et
al. applied an extended Fourier amplitude sensitivity analysis test to
reduce a 14-compartment whole body PBPK model of diazepam while
preserving the dynamical behaviour of the arterial concentration
compartment [47].