Abstract

Klebsiella pneumoniae is an important opportunistic pathogen associated with high levels of antibiotic resistance (AMR) and variable virulence capacity. Deciphering the link between AMR and virulence is a key objective in genomic epidemiology. For this project, we conducted a comparative genomics analysis of 45 K. pneumoniae isolates derived from an initial set of 84 strains after stringent quality control and sequencing. We generated the genome sequences using Illumina MiSeq technology. We constructed an AMR and virulence determinant binary feature matrix and subjected it to unsupervised clustering to identify structured genomic profiles. We identified four distinct clusters, corresponding to differences in the presence of resistance and virulence determinants. While most isolates showed moderate, independent distributions of these characteristics, some showed increased presence of both AMR and virulence determinants, indicating potential resistance-virulence convergence profiles. A highly discriminative (AUC=0.96) interpretable logistic regression model based on genomic burden features separated these genomes; however, the model is based on internal data and is exploratory. Partial concordance with 15 publicly available genomes further supports the biological feasibility of the resistance-virulence convergence paradigm. Overall, this work shows that Klebsiella pneumoniae has a heterogeneous genomic profile and presents an interpretable approach to identifying potential convergence profiles.