From 2010-12 I was a postdoc in the Modernising Medical Microbiology consortium, working on analysing whole genome sequences from bacteria. Most of my work was on Staphylococcus aureus. We studied the within-host diversity of S. aureus, when the bacteria is in asymptomatic carriage, and also when a strain in carriage goes on to cause disease. Further, we studied the population-wide diversity of S. aureus, with a particular focus on homologous recombination, in particular how this is affected by mobile elements. Another paper examined the extent to which antimicrobial resistance can be determined from the whole genome sequences of S. aureus.

Since leaving Oxford, I have continued to work on the genetics of human pathogens, but focussing on the computational aspects of Bayesian inference in this setting. This work has been funded by BBSRC, under the project "Understanding recombination through tractable statistical analysis of whole genome sequences". There have been two strands of work.

  • Sequential inference of phylogenetic trees. My group has been working on sequential Monte Carlo approaches to this problem, some of which are detailed in this paper. The video below gives a visualisation of this method in action. We are currently working on including this method in BEAST2.
  • Inference for models of recombination. Doing Bayesian inference for phylogenetic trees is hard: effective sample sizes from state-of-the-art MCMC algorithms are usually very low. When additionally considering recombination, the problem is significantly more complicated. The "ClonalOrigin" model offers some compromise between accuracy and computational tractability, but its inference (with reversible-jump MCMC) is still too expensive in most cases. Our work "Speeding up Inference of Homologous Recombination in Bacteria" describes the use of advanced reversible-jump MCMC methods for inferring this model. One of the talks I gave on this method can be found here.