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

Area of Science:

• Microbiology
• Genomics
• Machine Learning

Background:

• Vancomycin is crucial for treating Staphylococcus aureus infections.
• Increasing vancomycin resistance necessitates rapid diagnostic tools for vancomycin-intermediate S. aureus (VISA).

Purpose of the Study:

• To develop and validate a genome-based machine-learning model for discriminating between VISA and vancomycin-susceptible S. aureus (VSSA).

Main Methods:

• Utilized whole-genome sequences from 25 S. aureus isolates.
• Developed a machine-learning model incorporating 14 gene parameters, including typing markers and resistance-associated genes.
• Employed leave-one-out validation for model evaluation.

Main Results:

• The model achieved an 84% classification accuracy in distinguishing VISA from VSSA isolates.
• Discrimination was based on specific genomic markers of antibiotic susceptibility, not overall sequence relatedness.
• The model successfully differentiated isolates despite them not forming distinct evolutionary groups.

Conclusions:

• A genome-based machine-learning approach offers a generalized framework for classifying bacterial phenotypes.
• This method is valuable for identifying genomic features underlying antibiotic resistance.
• The developed model provides a tool for accurate VISA diagnosis.