Genomic neighbor typing for bacterial outbreak surveillance

Abstract

Genomic neighbor typing enables heuristic inference of bacterial lineages and phenotypes from nanopore sequencing data. However, small reference databases may not be sufficiently representative of the diversity of lineages and genotypes present in a collection of isolates. In this study, we explore the use of genomic neighbor typing for surveillance of community-associated Staphylococcus aureus outbreaks in Papua New Guinea (PNG) and Far North Queensland, Australia (FNQ). We developed Sketchy, an implementation of genomic neighbor typing that queries exhaustive whole genome reference databases using MinHash. Evaluations were conducted using nanopore read simulations and six species-wide reference sketches (4832 - 47616 genomes), as well as two S. aureus outbreak data sets sequenced at low depth using a sequential multiplex library protocol on the MinION (n = 160, with matching Illumina data). Heuristic inference of lineages and antimicrobial resistance profiles allowed us to conduct multiplex genotyping in situ at the Papua New Guinea Institute of Medical Research in Goroka, on low-throughput Flongle adapters and using multiple successive libraries on the same MinION flow cell (n = 24 - 48). Comparison to phylogenetically informed genomic neighbor typing with RASE on the dominant outbreak sequence type suggests slightly better performance at predicting lineage-scale genotypes using large sketch sizes, but inferior performance in resolving clade-specific genotypes (methicillin resistance). Sketchy can be used for large-scale bacterial outbreak surveillance and in challenging sequencing scenarios, but improvements to clade-specific genotype inference are needed for diagnostic applications. Sketchy is available open-source at: https://github.com/esteinig/sketchy