Multi-Scale Theoretical Investigations of Protein Interactions and Evolution

Abstract

Evolution of biological systems requires players of multiple layers, from atoms and molecules to organisms and populations. Expression of a gene is operated by molecular machineries for transcription, translation, regulation, and maintenance, which work in concert to produce certain macroscopic and observable phenotypes. And when these phenotypes are exposed to selective pressures, more fit phenotypes (with their genes, molecular machineries, and interaction networks) survive in the population. While the relationship of a gene to its cellular consequences is not fully elucidated, it is known that molecular interactions are one of the key factors that determine the relationship. In this dissertation, we introduce several theoretical tools to study protein interactions and evolution, and show their applications at various scales. The first tool is a coarse-grained scoring function that predicts binding free energy of a protein complex. The scoring function is a simple linear combination of exposed interface areas of different amino acids. In spite of the simplicity, it shows a reasonable predictive power, and predicts correct biochemistry qualitatively. The second is an analytical theory of a spin model on a simple graph, developed by using conventional statistical mechanics. We separated structural and energetic contributions to the free energy of the system, and also obtained a closed form of linear graph contributions. The closed form is applied to predict sequence space free energy of lattice proteins. Lastly, we introduce statistical methods to analyze cellular proteomes and transcriptomes. They can extract global responses of proteomes and transcriptomes to a perturbation, and also responses of specific gene groups. We applied the methods to E. coli and yeast systems to address questions on the genotype-phenotype relationship and evolution.