Multiscale Modeling in Systems Biology : Methods and Perspectives

Abstract: In the last decades, mathematical and computational models have become ubiquitous to the field of systems biology. Specifically, the multiscale nature of biological processes makes the design and simulation of such models challenging. In this thesis we offer a perspective on available methods to study and simulate such models and how they can be combined to handle biological processes evolving at different scales.The contribution of this thesis is threefold. First, we introduce Orchestral, a multiscale modular framework to simulate multicellular models. By decoupling intracellular chemical kinetics, cell-cell signaling, and cellular mechanics by means of operator-splitting, it is able to combine existing software into one massively parallel simulation.  Its modular structure makes it easy to replace its components, e.g. to adjust the level of modeling details. We demonstrate the scalability of our framework on both high performance clusters and in a cloud environment.We then explore how center-based models can be used to study cellular mechanics in biological tissues. We show how modeling and numerical choices can affect the results of the simulation and mislead modelers into incorrect biological conclusions if these errors are not monitored properly. We then propose CBMOS, a Python framework specifically designed for the numerical study of such models.Finally, we study how spatial details in intracellular chemical kinetics can be efficiently approximated in a multiscale compartment-based model. We evaluate how this model compares to two other alternatives in terms of accuracy and computational cost. We then propose a computational pipeline to study and compare such models in the context of Bayesian parameter inference and illustrate its usage in three case studies.