Authors
MacKenzie A P Wilke, Laura F Mataseje, Nicole Lerminiaux, Melissa McCracken, Romaine Edirmanasinghe, Melanie Baxter, Ken Fakharuddin, Gurman Grewal, Xiao Rui Li, Eric Marnier, Aaron Petkau, Michael Mulvey, George Zhanel, George R Golding, Amrita Bharat
Published in
Microbiology spectrum. Pages e0165226. Oct 06, 2026. Epub Oct 06, 2026.
Abstract
Treatment of healthcare-associated infections is becoming increasingly challenging due to rising antimicrobial resistance (AMR). As clinical microbiology and public health continue to transition toward whole-genome sequencing (WGS) for outbreak detection, diagnostics, and surveillance, the ability to accurately predict antimicrobial resistance from genomic sequence data will become crucial. In this study, in silico antimicrobial resistance prediction with the Staramr tool was refined from whole-genome sequencing data for four top-priority healthcare-associated bacteria: carbapenemase-producing Escherichia coli, carbapenemase-producing Klebsiella pneumoniae, methicillin-resistant Staphylococcus aureus, and vancomycin-resistant Enterococcus faecium. Isolates (n = 4,876) were collected in Canada within two national surveillance programs: the Canadian Nosocomial Infection Surveillance Program and the Canadian Ward Study. Broth microdilution was used as the gold standard to evaluate the performance of Staramr. The average sensitivity, specificity, and concordance of categorical antimicrobial resistance prediction compared to broth microdilution were 90.8%, 90.1%, and 92.4% for E. coli (n = 1,095); 87.7%, 91.7%, and 90.6% for K. pneumoniae (n = 687); 95.5%, 98.9%, and 96.3% for methicillin-resistant S. aureus (n = 2,150); and 95.4%, 97.9%, and 97.2% for vancomycin-resistant Enterococcus faecium (n = 944). Additionally, this internal tool refinement identified areas of improvement for antimicrobial resistance prediction. The average sensitivity, specificity, and concordance of antimicrobial resistance prediction were high (>90%) for most combinations of antimicrobials and species that were evaluated. While AMR prediction from WGS was very promising, it did not correlate well with current Clinical and Laboratory Standards Institute breakpoints for non-susceptibility for some antimicrobial-organism combinations. As in silico prediction of antimicrobial resistance has been improved in this study through tool refinement, it shows future promise for molecular methods to complement phenotypic antimicrobial susceptibility testing. This will allow for more precise tracking of the molecular epidemiology of resistance and cost savings when isolates are being sequenced for strain typing.
In silico prediction of antimicrobial non-susceptibility in priority healthcare pathogens is becoming an important area of study due to the rising rates of resistance and the transition to whole-genome sequencing. To evaluate whether Staramr, an antimicrobial resistance (AMR) prediction tool, can reliably detect non-susceptibility in four top-priority healthcare-associated bacteria, we performed a validation study on a large Canadian clinical data set, which included carbapenemase-producing Escherichia coli and Klebsiella pneumoniae, methicillin-resistant Staphylococcus aureus, and vancomycin-resistant Enterococcus. Compared to traditional antimicrobial susceptibility testing, we found that Staramr was reliable for predicting non-susceptibility for most combinations of antimicrobials and organisms; however, some combinations require additional data on genetic mechanisms of resistance for accurate prediction. Additionally, this study identified areas for refinement for current drug-gene interpretation keys. Thus, this tool refinement study advances the use of genotypic methods of AMR prediction for high-priority healthcare pathogens.
PMID:
42836344
Bibliographic data and abstract were imported from PubMed on 06 Oct 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 3
- Comments 0