Structured coalescent models for accurate Bayesian phylodynamic inference

Infectious disease events, especially those resulting from novel emerging pathogens, have significantly increased over the past few decades, possibly as a result of alterations in various environmental, biological, socioeconomic, and political factors (Chan et al., 2010). These factors - which include the increased global population, aging, expansion in international travel and trade, urbanization, and climate change - favour the emergence, evolution, and spread of new pathogens (Bloom et al., 2017), a trend that is very likely to continue. The term phylodynamics has been coined to describe infectious disease behaviour that arises from a combination of such evolutionary and ecological processes (Grenfell et al., 2004), and recent years have seen the emergence of various applications to perform phylodynamic inference. Central to these software packages are statistical and computational approaches that allow the reconstruction of the unobserved - and typically time-stamped - phylogeny relating to a molecular sequence sample, which serves as molecular epidemiologists' primary tool (Pybus and Rambaut, 2009), allowing to tackle key biological questions on viral epidemics. The objectives of my PhD are three-fold, and all contribute towards obtaining a thorough understanding of the evolution and spread of pathogens and developing novel tools and methodologies while doing so. Objective 1) A detailed review of the literature concerning phylogeographic analyses applied to the rabies virus (RABV). Rabies is a neglected zoonotic disease and approximately 60k people are still estimated to die from rabies each year, making RABV a pathogen of key interest in many phylogeographic and phylodynamic analyses that study its evolution and spread. I aimed to perform a thorough review of published studies on RABV in these research areas, making a distinction based on the geographic resolution associated with the available sequence data. I paid special attention to environmental factors that these studies found to be relevant to the spread of RABV and have highlighted a knowledge gap in terms of applying these methods when all required data were available but not (yet) fully exploited. Objective 2) Developing a visualization tool to easily interpret phylogeographic results. Phylogeographic analyses provide direct and actionable information, not only for researchers and public health agencies but also for practitioners in public health and epidemiology, who are instrumental in collecting pathogen samples for sequencing. Visualization tools permit researchers and health practitioners to explore the geographic range and sequence of infections through an easy-to-use graphical user interface. This allows the output from phylogeographic analyses to be translated into tangible responses. Objective 3) Develop an approach to perform Bayesian model selection of molecular clock models. In studying pathogen evolution, molecular clocks enable combining the genetic differences between samples and their collection times to estimate time-calibrated phylogenies. Along with the development of increasingly complex molecular clock models comes the need to determine which model is best suited to analyze a particular data set accurately. Employing a sub-optimal model may lead to inaccurate results. For this purpose, increasingly accurate marginal likelihood estimators have been developed in recent years to compare relative model fit in a Bayesian framework (Baele et al., 2016). Through a detailed simulation study, I will examine how these estimators can be used to determine the optimal molecular clock model for a given data set. I will pay special attention to molecular clock model comparisons that explore different phylogenetic search spaces.