ARIBA: rapid antimicrobial resistance genotyping directly from sequencing reads

Abstract

Antimicrobial resistance (AMR) is one of the major threats to human and animal health worldwide, yet few high-throughput tools exist to analyse and predict the resistance of a bacterial isolate from sequencing data. Here we present a new tool, ARIBA, that identifies AMR-associated genes and single nucleotide polymorphisms directly from short reads, and generates detailed and customizable output. The accuracy and advantages of ARIBA over other tools are demonstrated on three datasets from Gram-positive and Gram-negative bacteria, with ARIBA outperforming existing methods.

Keywords: antimicrobial resistance; bacteria; genotyping; sequence typing; whole genome sequencing.

Fig. 1.

Overview of the ARIBA mapping and targeted assembly pipeline.

Fig. 2.

Effect of read depth on the number of gene calls for all five van genes in the 17 vancomycin-resistant E. faecium samples.

Fig. 3.

Concordance between AMR calling methods on the S. sonnei data. A coloured dot indicates which methods were in agreement. For example, column 3 shows that there were 56 samples where ARIBA, SRST2 and Holt et al. [24] (Holt 2012) were in agreement in predicting the presence or absence of AMR. The first column illustrates where no resistance mechanisms were predicted.

Fig. 4.

Distribution of MICs (represented on a logarithmic scale) for AZM for all observed combinations of relevant AMR determinants in our custom database. Dotted horizontal lines mark clinical breakpoints. The lower line marks the lowest EUCAST (http://www.eucast.org/clinical_breakpoints/) breakpoint (0.25 µg ml⁻¹) and the upper line marks the post-2005 breakpoint used in the USA (2 µg ml⁻¹) [46].

Fig. 5.

Correlation between the number of alleles containing the 23S C2597T mutation (C2611T in Escherichia coli) in AZM-resistant isolates and their MIC values for this antimicrobial.