Project
Bayesian phylogenetic model development for viral evolutionary reconstruction
Statistical modeling is critical to capturing the complex processes involved in molecular evolution. A lot of attention has been devoted to identifying processes of natural selection at the molecular and how these may confound phylogenetic inference. In the first part of the project we will investigate the use of Markov modulated codon substitution models in characterizing processes of molecular adaptation. We will implement such models in a Bayesian statistical inference framework and assess their performance on simulated and empirical data. To address the computational burden of fitting such models to sequence data, we will explore the use of massively parallel likelihood computations on multi-core architectures. By complementing these models with stochastic mapping techniques, we aim to design new approaches to characterize site-specific substitution histories. In the second part of the project, we aim to develop flexible and computationally efficient Bayesian implementations of codon substitution processes that accommodate both among-lineage and among-site variation in synonymous and non-synonymous substitution rates. We will adopt Bayesian non-parametric models to flexibly accommodate heterogeneity in process of natural selection and to accurately estimate divergence times well past the point where the nucleotide substitution process would saturate. Also for these models, we will assess performance both from a statistical and computational perspective.