Scalable and Reproducible Modeling and Simulation for Heterogeneous Populations
Abstract
Advancements in the systems and synthetic biology fields have proved that biology can be engineered. The development of computer-aided design (CAD) tools has contributed to advancements in these fields. Mathematical modeling and simulation methods are important assets of CAD tools that are frequently applied to the systems and synthetic biology fields. Modeling and simulation methods are used to understand or predict the behavior of a biological system being studied. However, many modeling efforts in those fields face a reproducibility problem, where many published models are not reproducible. In order to address such issue, standards have been created for the representation of biological models. A major advantage of standards is that they enable model reuse and sharing. The leading standard representation of biological systems is the Systems Biology Markup Language (SBML). The SBML standard is used to describe how biological processes affect and modify biological entities in a system. Such standard has been widely used to describe biochemical networks, cell signaling path, and gene regulation, among others. Unfortunately, not all models use SBML since there are many biological systems that SBML is incapable of representing efficiently, such as heterogeneous cellular populations. This dissertation explores extensions to SBML for the efficient representation of large heterogeneous cellular populations and simulation methods that can simulate such complex models efficiently. Since cellular populations are inherently hierarchical, this dissertation proposes an efficient simulator for hierarchical SBML models. Since the hierarchical structure is preserved in the proposed simulator, the hierarchical simulator is a perfect fit for handling hybrid models. However, no one has explored the coupling of different modeling formalisms within the same SBML model. Hence, this dissertation proposes a methodology that can be used to describe hybrid models. Such methodology is demonstrated by using dynamic flux balance analysis (DFBA) models as examples and such models can be successfully exchanged between tools. This dissertation also discusses extensions to the SBML data model to support regular structures in the form of arrays. Arrays is well-suited for population models since population models use large regular structures. Another application of arrays is microsimulation of disease models, where a population of individuals with unique characteristics need to be model. With the proposed arrays extension, simulators need to scale in order to handle the increase complexity that the arrays extension introduces. Hence, this dissertation also proposes an efficient simulation method that takes advantages of arrays.
