Monte Carlo computational approaches in Bayesiancodon-substitution modelling
Nicolas Rodrigue, Nicolas Lartillot · Oxford University Press eBooks · 2012
This chapter reviews Markov Chain Monte Carlo (MCMC) approaches in codon-substitution modelling. It outlines the process of data analysis using the Bayesian framework. It describes the algorithms for sampling from posterior distributions under the more well-known models inspired from Muse and Gaut (1994) and Goldman and Yang (1994). The chapter also illustrates how parameter expansion and data-augmentation sampling methodologies can be employed to develop models that apply the Dirichlet process for capturing across-site heterogeneity (e.g. Huelsenbeck et al., 2006), and models that allow for dependence among codons due to protein structure (e.g. Robinson et al., 2003). It describes the thermodynamic integration approaches for computing Bayes factors in these contexts, and highlights the diversity of models that can be implemented and directly compared with such approaches.