The study of phylodynamics uses the information in pathogen genomes to understand and predict epidemic dynamics and outbreaks. We use epidemiological techniques coupled with the latest phylogenetic models to study problems of national and international importance.

A unified framework for phylodynamic inference of infectious diseases

To reduce the burden of infectious diseases, we must understand how diseases spread so we can best intervene to stop them. The most promising new tools are fast, accurate and cheap genome-sequencing technologies which provide a flood of potentially informative data about how, when and where diseases spread. The necessary techniques to interpret this valuable data, however, are not yet available. 

This project aims to produce a flexible yet practical framework for conducting phylogentics-based inference under sophisticated epidemiological models. The two key objectives are: 

  1. Develop and implement a model of the transmission tree that respects a stochastic model of epidemic dynamics within a structured population
  2. Develop and implement a model for the pathogen gene tree conditional on the transmission tree
This programme of work will develop a coherent hierarchy of models, progressively including greater levela of detail to accurately model transmission trees, the pathogen population dynamics and the mutational bottleneck that occurs when a pathogen is transmitted from one host to another. 
Researchers: Dr Tim Vaughan, Alexandra (Sasha) Gavryuskina, Dr David Welch, Prof Alexei Drummond

Modelling epidemic curves


The number of people infected by an outbreak through time is known as the epidemic curve.  It can be thought of as the size of the viral population through time.  Since 2005, it has been possible in BEAST to estimate population sizes through time from sampled genomic sequences http://mbe.oxfordjournals.org/content/22/5/1185.full.  But these methods haven't taken into account the peculiar nature of epidemic curves which are well described by mathematical models.  We have developed a number of methods that can approximately estimate the epidemic curve (eg http://www.pnas.org/content/110/1/228.long) nad epidemiological constants such as the reproduction number of r an outbreak, R0.

Recently, we have developed a method that can make very accurate estimates of the epidemic curve and the parameters for a range of commonly used compartmental epidemic models.  The goal of this work is not just to recover the epidemic curve but to let the epidemic curve inform the phylogenetic model to come to a more detailed understanding of epidemic circulation.  Preprint due out shortly!

Researchers: Dr Tim Vaughan, Dr David Welch


Inferring pattern of global influenza migration using MultiType Tree package


Understanding exactly where infections come from and how they mix across the globe is central to understanding how to control them.   Modern epidemics are a mix of global and local outbreaks: local because we spread infection to those nearby but also global due to an ever more mobile population.

In population genetics, we use the structured coalescent to understand this sort of set-up: here the global population is divided into demes where individuals mix freely, while mixing between demes can occur but is less frequent. We can think of each deme being associated with a colour, so that when we draw the phylogenetic tree for a sample of individuals, the location of the ancestral lineages is shown by the colour of the lineage. While the theory is reasonably well-understood, actually using this model for inference is notoriously difficult.  

We have put together the multi-type tree package in Beast2 that allows inference under the structured coalescent to be coupled with all the standard Beast features such as variable population size and the full range of substitution and clock models. The power of this tool is shown here in an analysis of H3N2 influenza virus with samples from New Zealand, Hong Kong and New York.

Paper: Vaughan, T. G., Kühnert, D., Popinga, A., Welch, D., Drummond A. J. (2014) Efficient Bayesian inference under the structured coalescent Bioinformatics 30 (16): 2274-2279 doi: 10.1093/bioinformatics/btu201

Researchers: Dr Tim Vaughan, Dr Denise Kühnert, Alex Popinga, Dr David Welch, Prof Alexei Drummond