Evolutionary insights into protein structure, stability, and functionality

The goal of this thesis work was the computational simulation of the evolution of populations of lattice-model and off-lattice model protein sequences to study (a) proteins as a general class of molecules, (b) how selective pressures, constrained by the existence and accessibility of properties under selection and affected by various mutational processes, result in the properties of modern sequences, and (c) the accuracy of sequence reconstruction algorithms. While the application of simple protein models towards most of these ends is not new, their easily controllable nature provides customizable approaches towards the many unique challenges imposed by the complexity inherent in biological systems, and the many previous uses has demonstrated their appropriateness. The model used for the majority of this thesis work included several novel features: exhaustive enumeration of all possible protein structures for a thermodynamic simulation of protein folding (involving all twenty common amino acids), a ligand-binding model for the study of selection for functionality, and a function relating binding probability to fitness based on Michaelis-Menten kinetics. Work presented in this thesis has covered a variety of topics. The study of how population evolution on the vast, multi-dimensional space of protein sequences can affect the distribution of three-dimensional structures revealed that historical accidents can greatly effect on subsequent evolution, and that the influence of pressures for functionality may be limited compared to the pressure for stability. Work examining the reasons why the majority of proteins are only marginal stable in a compact structure has underscored the importance of the distribution of properties across this sequence space on the properties that evolve, namely that a limited repertoire of highly stable sequences is sufficient explanation for the prevalence of marginal stability, and that adaptationist explanations regarding functionality are not required. Work examining the fitness effects of mutation and recombination of protein sequences has shown that even conservative homologous protein recombination can provide an evolutionary advantage to populations. Finally, evolution simulations performed using much larger, off-lattice proteins to generate known phylogenies have led to the discovery that Bayesian methods of phylogenetic reconstruction may be most accurate at estimating properties of extinct protein sequences.