Download files
Abstract
Measuring selective pressures shaping the evolution of viral populations is important for preventing and controlling the spread of disease, as well as for understanding evolutionary processes in general. Traditional methods for detecting and quantifying selection assume that a single segregating allele is under constant selection in a population of constant size. However, viruses frequently violate these assumptions due to (i) their high mutation rates and (ii) their complex epidemiological dynamics. We examine the effect of these factors using computational models describing evolution at protein-coding regions, under various population dynamics. In Chapter 2 we show, assuming population sizes are constant, that linkage-induced interference between segregating mutations distort commonly used statistics such as dN/dS and the McDonald-Kreitman (MK) statistic. We propose three alternative statistics to detect the effect of background selection, hitch-hiking and clonal interference. In Chapter 3, we examine selection acting in the context of an epidemiological multi-strain SIRS model, by explicitly modelling the effect of cross-immunity between related strains, competition for susceptible hosts, and decaying host immunity. By studying the probability of antigenic reversion, we show that time-varying antigenic selection mediated by host immunity has a qualitatively different effect from constant selective constraint, which is observable from changing frequencies at antigenic sites over time. In Chapter 4, we apply both of these methods to avian influenza, demonstrating their utility in comparing selection between different lineages. In combination, these methods allow us to distinguish between different forms of selection, which may allow us to discriminate between potential biological mechanisms shaping viral populations.