Benchmarking of tools for resolving plasmids from short-read assemblies for Klebsiella pneumoniae

The critical role of plasmids, particularly in the dissemination of AMR and virulence in nosocomial pathogens like Klebsiella pneumoniae , underpins the need for robust and scalable tools for plasmid identification and reconstruction from large-scale datasets of short-read sequence data that are routinely generated in clinical and public health settings. Here, we sought to evaluate all available tools to determine which are best suited for reconstructing the plasmidome of K. pneumoniae and related species from the species complex (KpSC). We used a publicly available dataset of 568 diverse KpSC isolates that had high-quality short-read Illumina data and matched complete genomes generated from hybrid assemblies incorporating additional long-read sequence data. This allowed us to investigate which tool perform best at recovering plasmid sequences when only short-read data are available.

None of the tools tested offered a comprehensive or consistently reliable solution to assembling plasmids from short-read sequence data of KpSC. Among the six tools that were benchmarked, performance varied across total runtime, RAM usage, prediction accuracy and sensitivity. Future tools developed in this space should offer meaningful advancements over existing tools and be rigorously evaluated using large, standardised bacterial datasets that reflect the diversity and complexity of plasmid content to ensure comparability across benchmarking studies.