Continue-tijd Markov keten modellen voor fylodynamica van pathogenen: model extensie, simulatie en visualisatie

The emergence of pandemic outbreaks, which we w itness with an alarmingly increasing frequency, re present a clear and imminent threat to the public health. This threat has been exemplified by the H1N1 influenza A outbreak in 2009 or¨ the Ebola virus outbreak in West Africa in 20 14. Combating viral spread and their associat ed disease burden is a tremendous challenge requir ing significant research efforts and decided¨ public health measures. Viral sequence data, ¨which is now becoming available in unprecedented¨ quantities and with remarkable rapidity, is f requently considered to be one of the main assets¨ in the characterization of pathogens associated wi th high mortality. The research direction coi ned `phylodynamics&##39; integrates molecular and e pidemiological data and aims to develop a fully qu antitative understanding of the processes tha t shape the epidemiology, evolution and spread of¨ viruses. In this thesis I outline my co ntributions to pursue comprehensive analyses¨ of the interaction between the ecological and evol utionary dynamics of emerging viruses. In par ticular this PhD project aims at advancing statist ical models for unravelling viral spatiotempo ral history, to provide adequate visualization of¨ our statistical estimates and to alleviate the com putational burden associated with drawing inf erence under these models. I believe that the se developments have the potential to result¨ in a better understanding of pathogen epidemi ology and evolution, which may ultimately ass ist in designing effective intervention and preven tion strategies. Chapter 1 of this thesis in troduces fundamental aspects of phylogenetics . Section 1.1 discusses what processes shape¨ the evolution and emergence of pathogens and how p hylogenetics inference can be used to infer those¨ past processes. In Section 1.2 I then proceed ¨to introduce the theory of Markov chains and ¨likelihood computation, and how this theory¨ is used to formulate stochastic models of sequence ¨evolution. Simple examples are included to illustrate these concepts. Finally Section 1.4 ¨covers Bayesian methods and MCMC algorithms, and¨ provides some detail of their implementation¨ in the BEAST software. Chapter 2 is an expan ded version of our published manuscript on time-he terogenous ‘epoch’ models [Bielej ec et al., 2014]. I included details of the i mplementation of the numerical procedures in BEAST ¨software package and BEAGLE library, as well ¨as speed-up comparisons between sequential (CPU)¨ and parallel (GPU) implementations, which wer e omitted in the published version of the manuscri pt. This chapter is accompanied by Appen dix A, which describes how the implemented models¨ can be setup as an XML document readable by BEAST. In Section 5.1 I discuss interesting avenues ¨for further extensions and ongoing work on this model. Chapters 3 and 4 are publish ed manuscripts in which I introduce two softw are packages written and maintained as part o f my PhD training. Specifically, Chapter 3 pr esents software for visualizing phylogeographic re constructions resulting from Bayesian inference of viral diffusion processes called SPREAD.This particular program has proven to be qu ite popular, which is why in Section 5.3 I discuss ¨extensions and improvements which are planne d to appear in the second version of the package. A detailed tutorial on using the cu rrent version of this package can be found in Appe ndix B. The πBUSS simulation software pac kage is discussed in Chapter 4, along with a simul ation study that aims at exploring the limits of e stimating old divergence times when faced wit h molecular data coming from rapidly evolving viru ses. This study has sparked my interest¨ in developing more biologically realistic codon mo dels, which allow for both across-site and across- branch variation. Such models should improve the r esolution with which we can estimate old ¨divergence dates. This ongoing work is descr ibed in Section 5.2. Future extensions of the ¨πBUSS simulator are described in Section¨5.4 and a gentle tutorial on using this softw are can be found in Appendix C. Finally Chap ter 5 offers a wider perspective on the presented¨ work. In Subsection 5.3.1 of this c hapter I describe some of the past and curren t efforts to visualize spatio-temporal diffusion p rocesses, and how SPREAD is positioned in relation ¨to those endeavors. In the last Section 5.5 I briefly touch upon the history of high-perf ormance computing and proceed to discuss some of t he parallel computing architectures which are ¨in use today. These advances are of utmost i mportance for the data-driven bioinformatics commu nity, and learning how to properly lever age their computing capabilities is a necessity if ¨we want to capitalize on the vast amount of genomic data collected every day.status: publishe