The covarion model of molecular evolution : a thesis presented in partial fulfilment of the requirements for the degree of Master of Philosophy in Biology at Massey University

Abstract

Current methods for constructing evolutionary trees generally do not work well for sequences in which multiple substitutions have occurred. The covarion hypothesis may provide a solution to this problem. This hypothesis states that only a limited number of the codons in a given sequence are free to vary, but that the set of variable codons may change as mutations are fixed in the population. Although this is reasonable from a biological point of view, it is a difficult hypothesis to test scientifically because the apparent large number of parameters involved makes it very hard to analyse statistically. In this study, computer simulations were carried out on up to 51 machines running in parallel, using a simple covarion model based on a hidden Markov model (HMM) approach. This model required two new parameters—the proportion of sites that are variable at any given time, and the rate of exchange between fixed and variable states. These two parameters were both varied in the simulations. Sequence and distance data were simulated on a given tree under this covarion model, and these data were used to test the performance of standard tree-building methods at recovering the original tree The neighbour joining and maximum likelihood methods tested were found to perform better with data generated under the covarion model than with data generated under a simpler model in which all sites vary at the same rate. This suggests that current tree-building methods may perform better with biological data than computer simulation studies suggest.