RAPID INFERENCE OF ANTIBIOTIC RESISTANCE AND SUSCEPTIBILITY BY GENOMIC NEIGHBOR TYPING
Surveillance of drug-resistant bacteria is essential for healthcare providers to deliver effective empiric antibiotic therapy. However, traditional molecular epidemiology does not typically occur on a timescale that could impact patient treatment and outcomes.
We developed a method called ‘genomic neighbor typing’ for inferring the phenotype of a bacterial sample by identifying its closest relatives in a database of genomes with metadata. We implemented this with rapid k-mer matching, which can be used on Oxford Nanopore MinION data and run in real time. We adopted the method for pneumococcus using a database of 616 genomes from a carriage study in Massachusetts children and five antibiotics. Finally, we evaluated the method using isolates and metagenomes from geographically distinct regions.
We show that genomic neighbor typing can infer antibiotic resistance and susceptibility of S. pneumoniae isolates within five minutes of sequencing starting (sens/spec 91%/100%) and for clinical metagenomic sputum samples within four hours of sample collection (75%/100%). We also show how the method can be adopted for custom species and drugs.
Genomic neighbor typing has wide application to pathogen surveillance and may be used to greatly accelerate appropriate empirical antibiotic treatment.