D.R.EVOL: Three Dimensional Realistic Evolution

Abstract

Simplifying assumptions are necessary to model complex biological processes. Although some assumptions may make sense mathematically, they are often implausible when literally translated. This is especially true of the independence among codons assumption, which states that the evolutionary rate at one codon isindependent of the evolutionary rate at surrounding codons. Sites within proteinsmust interact in order to form intricate three-dimensional binding sites and activation domains. This dissertation details the derivation of a procedure for statistical inference when independent change is not assumed.

The procedure is implemented in a Bayesian framework where Markov chain Monte Carlo methods permit approximation of posterior distributions. Analyses with the procedure on data sets with two and three taxa are explored and biologically plausible values of the solvent accessibility and pairwise interaction parameters are inferred. Via these analyses, we illustrate the chronological ordering of amino acid replacements and the detection of specific events to be positively selected. We also find spatial clustering of the amino acid replacements that have most affected sequence-structure compatibility during the evolution of primate eosinophil-derived neurotoxin proteins.