Macromolecular Self-Assembly: Simulation and Optimization

This thesis develops computational methods for the investigation of self-assembly systems in biology as well as methods for the simulation of reaction-diffusion chemistry. We discuss the current state of the field with respect to modeling self-assembly and its importance to systems biology generally. Our contributions come in the form of pipelines for model inference based on comparisons of in silico experiments with physical experiments monitoring assembly progress. A new black-box parameter optimization methodology suitable for noisy objective values, and using multiple Gaussian processes, is presented. We also discuss the current landscape of course-grained simulation methods for reaction-diffusion chemistry and their limitations. A novel algorithm generalizing the stochastic simulation algorithm to continuous space is presented. We describe its physical justification as well as its improvements over the state of the art in certain respects, e.g. run time efficiency. At the end, we describe our applied work in collaboration with the Faeder and Murphy Labs (University of Pittsburgh and CMU, respectively) on an immune cell signaling project. While not directly related to self-assembly or the methods described previously, this collaboration allowed us to design a kinetic model from scratch and develop an optimization framework tailored to real experimental data.