Overcoming Complexity in Systems Biology Modeling and Simulation

Advances in high-precision and high-throughput biochemical measurement have generated overwhelming amounts of molecular and cellular data. To make sense of this biological information, computational approaches for building and exploring models of cellular function are needed. Here I present methods and software that confront three of the most pressing computational challenges facing biological modeling: 1) managing the enormous numbers of possible molecular interactions, 2) coarse-graining complex reaction mechanisms in a flexible and efficient manner, and 3) merging multiple biological representations and simulation methods to construct multiscale models. The result is two new computational modeling platforms: NFsim and the Hive. NFsim is a biochemical reaction simulator which uses a rule- and agent-based approach to simulate the dynamics of large biochemical reaction networks. NFsim also addresses difficulties in coarse-graining by providing the capability to incorporate steady-state approximations, molecular cooperativity, and logical constraints in kinetic models. The Hive is a scalable software framework designed for constructing multiscale biological models. Its key feature is a general interface for merging multiple biological representations and simulation methods, for instance, to couple a stochastic intracellular signaling model with partial differential equations describing extracellular diffusion. I demonstrate the advantages of NFsim and the Hive across a wide range of biological systems, including signaling in the immune system, polymerization of actin filaments, genetic regulatory networks, and microbial signaling pathways. The culmination of this thesis is an extended analysis of the effects of multiple flagellar motors and molecular noise on bacterial chemotaxis. By using experimentally calibrated models of chemotactic bacteria developed with NFsim and the Hive, I illustrate how signaling noise in the pathway is predicted to coordinate the switching statistics of multiple flagellar motors and improve chemotactic response to shallow gradients of attractants. This counter-intuitive, noise-enhanced response is possible because noise not only affects signal processing, but also modifies the motion of a bacterium and the statistics of gradient sampling. These results have important general implications for how noise influences other signal transduction and gradient detection systems.