Predicting the emergence of host-adapted bacterial phytopathogens

Using low cost, high throughput genome sequencing we will ask how the population structures of Ps lineages on cultivated and wild cherry varies over time, and how much this is shaped by host genotype and local environment. Profiling absolute levels of abundance of bacterial and fungal species in the phylloplane we will ask how agronomic practice, specifically the application of nitrogen and the use of polytunnel covers affects epiphytic and pathogenic lineages in realistic field settings. Convergent effector gain has been identified in pathogens of cherry and we will examine if effector repertoire is also adapted to colonisation of the shoot surface. We will carry out controlled evolution experiments to study whether effector rich lineages are able to colonise increasingly phylogenetically distant hosts through and whether the adaptive potential of effector poor, toxin-rich Ps lineages to shift hosts is greater than effector-rich lineages. Using phage transfer experiments under different stress-inducing conditions we ask whether there is phage-mediated effector transfer between pathogens and epiphytes as is indicated by preliminary work. Utilising the abundance of Ps genomic data, coupled with supervised machine learning approaches, we will develop tools to predict bacterial host range from genome sequence alone. Developing a training set of genomes with well established host-pathogen association, then testing a range of feature classification techniques and machine learning method, we will evaluate predictions of host compatibility with Prunus and other plant species. Using 'deep learning' methods, we will discover further explanatory features associated with host range classification. We will validate predictions through pathogenicity tests on predicted compatible hosts.