Simulation and optimization of logical and kinetic biochemical models

Abstract

During the last years the impact of systems biology has grown drastically. In contrast to traditional biology, this interdisciplinary field comprises the investigation of biological processes from a systems perspective. An example for a systems biology project is the Virtual Liver Network, in which metabolic liver function is modeled computationally. A mathematical model in systems biology provides a hypothesis that is testable by biological experiments. Data obtained from model simulation can thereby be compared to experimental data possibly leading to the adaptation of the model. Experimental data is also used to optimize a model, e.g., to estimate certain model parameters or to identify connections between model components.

From the different kinds of computational models used in systems biology logical models and kinetic ordinary differential equation (ODE) models are covered in this thesis. While logical models enable to describe biological processes qualitatively, kinetic ODE models allow the dynamic description of these processes. Methods for the simulation and optimization of both model types were developed and applied here.

The first part of the thesis contains the application of a specific logical modeling technique called fuzzy logic modeling. A previously published method based on prior knowledge and experimental data was adapted to identify regulatory events responsible for the downregulation of drug metabolism during inflammation. Further experiments backed the hypothesis suggested by the model. The respective study conducted in collaboration with biologists is a relevant part of the Virtual Liver Network.

In the following part of the thesis an algorithm for the simulation of models given in the Systems Biology Markup Language (SBML) is described. SBML is the most important standard for storing and exchanging systems biology models. It enables to describe ODE models that can also contain other elements, such as rules for model components and events representing sudden changes of components. Because of these additional elements, simulation of SBML models is difficult and only few software tools support this standard completely. The Systems Biology Simulation Core Library (SBSCL), which contains an implementation of the developed algorithm, supports SBML completely. Benchmark tests were used to prove the correctness of the library. The SBSCL can be easily integrated into larger software tools. One example for this is SBMLsimulator, which connects the SBSCL and the optimization toolbox EvA2 and is described in the following chapter of this thesis. SBMLsimulator enables simulation and optimization of SBML models and can be very helpful in systems biology studies.

At the end of the thesis SBMLsimulator is applied in a large study investigating the influence of experimental noise on the estimation of kinetic parameters for three published SBML models. The study shows the usefulness of the tool, but also suggests to take the noise into account during estimation, which was previously only tested for a small toy model. Furthermore, the chosen approach is a way of robustness analysis that can be used to estimate how reliable the parameter estimates for a certain model are. Like the other methods applied and developed in this thesis, it can be used for further systems biology research.