Analysis of Multidrug Resistance in Staphylococcus aureus with a Machine Learning-Generated Antibiogram

Area of Science:

• Microbiology
• Computational Biology
• Epidemiology

Background:

• Multidrug resistance (MDR) surveillance requires understanding prevalence and phenotypes.
• Identifying specific associations underlying MDR patterns is crucial for effective antimicrobial stewardship.

Purpose of the Study:

• To apply machine learning and graphical networks to quantify and visualize associations between resistance traits in Staphylococcus aureus isolates.
• To identify and track clinically relevant multidrug resistance patterns.

Main Methods:

• Analysis of 1,091 Staphylococcus aureus isolates (2008-2018) using antimicrobial susceptibility testing.
• Application of association mining to identify and quantify relationships between two or more antimicrobial resistance traits.
• Comparison of resistance patterns based on year, methicillin susceptibility, and infection site.

Main Results:

• Identified common MDR phenotypes and confirmed known pairwise AMR trait relationships.
• Observed varying strengths of associations between beta-lactams and other antimicrobial classes across infection sites and methicillin susceptibility.
• Quantified specific co-occurrence of resistance traits, e.g., clindamycin and tetracycline resistance in skin infections.

Conclusions:

• Association mining efficiently discovers and quantifies AMR trait associations in Staphylococcus aureus.
• This approach enables comparison of resistance patterns across isolate subsets for tracking clinically relevant MDR.
• Findings can inform antimicrobial stewardship programs by highlighting specific, actionable resistance patterns.