A new method for identifying site-specific evolutionary rates and its applications.

In this thesis, I discuss each stage in the development of a new method for identifying site specific evolutionary rates, from conception of the idea, through the implementation to its application to data. TIGER, or tree independent generation of evolutionary rates, is based largely around the works of LeQuesne (1989), Wilkinson (1998) and Pisani (2004) and the premise that sites in a multi-state character matrix could be scored based on the level of agreement it displays with the other sites. In these earlier studies, however, agreement was measured in binary manner: sites were either compatible with each other or they are not. TIGER allows various degrees of agreement to occur between two sites, allowing it to pick up more subtle signals in the data. After implementing the method into a software program, it could be applied to data. Using a combination of simulated and empirical datasets, TIGER was shown to produce desirable results. In particular, removal of sites identified by TIGER was shown to improve phylogenetic reconstruction of deeply diverging lineages and of taxa displaying compositional attraction. Additionally, TIGER was applied to a gene content matrix in order to identify HGT signals and integrated into the analysis of a current phylogenetic problem, the origin of the mitochondria. Although it is widely accepted that eukaryotes have a chimeric genome, the specific “parent” of the mitochondria is, as of yet, unclear. Previous studies have failed to reach agreement regarding this issue for a number of reasons. Exploration of the signals using TIGER and heterogeneous modelling reveal that multiple signals and compositional heterogeneity are among the biggest problems with datasets containing both mitochondrial and a-proteobacterial sequences