Multi-objective Optimization with modeFRONTIER Applied to Systems Biology

Adam Thor (EnginSoft Nordic) Elin Nyman (Linköping University)

In this article modeFRONTIER was used to automate and optimize an analysis model written in MathModelica, a modelling and simulation software based on Modelica. The goal of the optimization process was not to identify a single solution to the model-fitting problem, but rather multiple solutions with acceptably small errors but at the same time with as widely varying parameters as possible.


?A significant number (tens of thousands) of MathModelica simulations were run through modeFRONTIER and several thousand solutions with an acceptably small error between measurement data and model predictions were identified. Since the goal was to identify different sets of solutions, a Partitive Clustering Analysis was carried out on the data.

By using Partitive Clustering Analysis, one of the tools available in modeFRONTIER for Multivariate Analysis (MVA), information regarding complex system behaviour that could not readily be understood using the normal tools available in the Design Space, such as Scatter Charts and Parallel Charts, was identified.

For an animated explanation of insulin signalling, watch the movie below.