Public defense of the dissertation of Mr. M.Sc. Till Köster on the topic: "Efficient Abstraction and Execution of Stochastic Simulation Models“

Stochasticity is found throughout cell-biological models when describing phenomena such as diffusion and the resulting random interactions of entities. This thesis is concerned with the efficient execution of these models. We cover the fundamentals modeling from the bottom up. Reaction systems are a main topic, particularly those that use Stochastic Simulation Algorithms. We discuss efficient variations of these algorithms, including their parallel execution on CPUs and GPUs. Furthermore, the limitations of static reaction systems will be overcome by introducing dynamic compartmentalization, which is vital for many processes. The arising challenges in efficient execution (including of a spatial particle variant) are addressed through novel algorithms and implementations. Another focus will also be the technical realization of the simulation abstraction through domain-specific languages and their interplay with simulator efficiency. This includes an application of the cell-biological simulation methods to agentbased models. Finally, we consider the issues of replication and evaluation of simulator implementations, in particular, as they relate to the questions of stochasticity and determinism. These introductions lay the groundwork for the six publications that comprise this thesis’s core. The articles present aspects of this thesis in more detail.