Taking the next-gen step: comprehensive antimicrobial resistance detection from Burkholderia pseudomallei genomes

Antimicrobial resistance (AMR) is emerging as a major threat to human health worldwide. Whole-genome sequencing (WGS) holds great potential for rapidly and accurately detecting AMR from genomic data in the diagnostic laboratory setting. However, most work to date has focussed on identifying only horizontally-acquired AMR-conferring genes, with chromosomally-encoded AMR determinants remaining largely undetected. Here, we present an improved tool for Antibiotic Resistance Detection and Prediction (ARDaP) from WGS data. ARDaP was designed with three priorities: 1) to accurately identify a wide range of AMR genetic determinants (i.e. horizontally-acquired gene gain, single-nucleotide polymorphisms, insertions-deletions, copy-number variation, and functional gene loss); 2) to predict enigmatic AMR determinants based on novel mutants with moderate- or high-consequence impacts in known AMR-conferring genes, and 3) to detect minor AMR allelic determinants in mixed (e.g. metagenomic) sequence data. ARDaP performance was demonstrated in the melioidosis pathogen, Burkholderia pseudomallei, due to its exclusively chromosomally-encoded AMR determinants and inherently limited treatment options. Using a well-characterised collection of 1,063 clinical strains, ARDaP accurately detected all currently known AMR determinants in B. pseudomallei (~50 determinants), including stepwise AMR mutations. Additionally, ARDaP accurately predicted meropenem resistance in four previously uncharacterised B. pseudomallei isolates. In mixed strain data, ARDaP identified AMR determinants down to ~5% allelic frequency, enabling the early detection of emerging AMR. We demonstrate that ARDaP is an accurate tool for identifying and predicting all confirmed B. pseudomallei AMR determinants from WGS data, including from mixed strain data. Finally, our study illustrates that manual cataloguing and functional verification of putative AMR determinants in individual pathogens is essential for truly comprehensive AMR detection. ARDaP is open source and available at: https://github.com/dsarov/ARDaP