Discordance between different bioinformatic methods for identifying resistance genes from short-read genomic data, with a focus on Escherichia coli

Several bioinformatics genotyping algorithms are now commonly used to characterize antimicrobial resistance (AMR) gene profiles in whole-genome sequencing (WGS) data, with a view to understanding AMR epidemiology and developing resistance prediction workflows using WGS in clinical settings. Accurately evaluating AMR in Enterobacterales, particularly Escherichia coli, is of major importance, because this is a common pathogen. However, robust comparisons of different genotyping approaches on relevant simulated and large real-life WGS datasets are lacking. Here, we used both simulated datasets and a large set of real E. coli WGS data (n=1818 isolates) to systematically investigate genotyping methods in greater detail. Simulated constructs and real sequences were processed using four different bioinformatic programs (ABRicate, ARIBA, KmerResistance and SRST2, run with the ResFinder database) and their outputs compared. For simulation tests where 3079 AMR gene variants were inserted into random sequence constructs, KmerResistance was correct for 3076 (99.9%) simulations, ABRicate for 3054 (99.2%), ARIBA for 2783 (90.4%) and SRST2 for 2108 (68.5 %). For simulation tests where two closely related gene variants were inserted into random sequence constructs, KmerResistance identified the correct alleles in 35 338/46 318 (76.3 %) simulations, ABRicate identified them in 11 842/46 318 (25.6 %) simulations, ARIBA identified them in 1679/46 318 (3.6%) simulations and SRST2 identified them in 2000/46 318 (4.3 %) simulations. In real data, across all methods, 1392/1818 (76%) isolates had discrepant allele calls for at least 1 gene. In addition to highlighting areas for improvement in challenging scenarios, (e.g. identification of AMR genes at <10x coverage, identifying multiple closely related AMR genes present in the same sample), our evaluations identified some more systematic errors that could be readily soluble, such as repeated misclassification (i.e. naming) of genes as shorter variants of the same gene present within the reference resistance gene database. Such naming errors accounted for at least 2530/4321 (59 %) of the discrepancies seen in real data. Moreover, many of the remaining discrepancies were likely 'artefactual', with reporting of cut-off differences accounting for at least 1430/4321 (33%) discrepants. Whilst we found that comparing outputs generated by running multiple algorithms on the same dataset could identify and resolve these algorithmic artefacts, the results of our evaluations emphasize the need for developing new and more robust genotyping algorithms to further improve accuracy and performance.