Population Analysis of Pseudomonas aeruginosa Virulence

Pseudomonas aeruginosa (PA) causes frequent and severe infections in hospitalized patients. In addition, the prevalence of multidrug-resistant PA is increasing and is now between 15-30% in many areas. Thus, it is not surprising that the IDSA, WHO, and the CDC have each listed PA as serious public health priority in need of new therapeutic agents. An age-old question concerning PA is why some strains cause substantially more aggressive infections than others. The recent application of next generation sequencing platforms to this problem has begun to provide an explanation by demonstrating that PA genomes differ substantially from strain to strain. Approximately 10-15% of the genes in a typical PA strain are "accessory," meaning that they are present in some strains but not others. Likewise, the "core" or conserved genome contains numerous single nucleotide variants (SNVs) and small insertion-deletions (indels). Although a few of these accessory genes and core alleles have been characterized and shown to modulate virulence, they represent the "tip of the iceberg." A systematic examination of strain-to-strain differences in PA is likely to uncover a wealth of novel virulence-impacting genes and alleles. Identification of these would have several important consequences: (i) They would dramatically enhance our understanding of PA virulence and the mechanisms by which this bacterium causes severe disease; and (ii) they would allow one to predict the virulence of PA strains based on the complement of accessory and core virulence genes/alleles that were present in their genomes. We hypothesize that application of comparative genomic approaches to large numbers of PA isolates will identify novel virulence genes/alleles and allow the generation of machine learning models to predict the virulence of PA isolates based on their genomes. We will perform the following specific aims to test these hypotheses: (1) Use machine learning models to predict the virulence of PA isolates based upon their genomic content. (2) Identify accessory genes and core genome SNVs/indels that play a causal role in virulence. (3) Develop a genome-based model that predicts clinical outcomes in patients with PA bloodstream infections. The impact of this proposal is twofold. First, it will lay the foundation for future models that provide valuable prognostic information to clinicians treating PA-infected patients. Second, it will identify new PA virulence factors that mediate novel pathogenic mechanisms of infection.