Format

Send to

Choose Destination
J Bacteriol. 2014 Mar;196(5):940-8. doi: 10.1128/JB.01410-13. Epub 2013 Dec 20.

Genome sequence-based discriminator for vancomycin-intermediate Staphylococcus aureus.

Author information

1
School of Biology, Georgia Institute of Technology, Atlanta, Georgia, USA.

Abstract

Vancomycin is the mainstay of treatment for patients with Staphylococcus aureus infections, and reduced susceptibility to vancomycin is becoming increasingly common. Accordingly, the development of rapid and accurate assays for the diagnosis of vancomycin-intermediate S. aureus (VISA) will be critical. We developed and applied a genome-based machine-learning approach for discrimination between VISA and vancomycin-susceptible S. aureus (VSSA) using 25 whole-genome sequences. The resulting machine-learning model, based on 14 gene parameters, including 3 molecular typing markers and 11 genes implicated in reduced vancomycin susceptibility, is able to unambiguously distinguish between the VISA and VSSA isolates analyzed here despite the fact that they do not form evolutionarily distinct groups. As such, the model is able to discriminate based on specific genomic markers of antibiotic susceptibility rather than overall sequence relatedness. Subsequent evaluation of the model using leave-one-out validation yielded a classification accuracy of 84%. The machine-learning approach described here provides a generalized framework for the application of genome sequence analysis to the classification of bacteria that differ with respect to clinically relevant phenotypes and should be particularly useful in defining the genomic features that underlie antibiotic resistance.

PMID:
24363339
PMCID:
PMC3957707
DOI:
10.1128/JB.01410-13
[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for HighWire Icon for PubMed Central
Loading ...
Support Center