A risk assessment framework for multidrug-resistant Staphylococcus aureus using machine learning and mass...

Area of Science:

• Microbiology
• Infectious Diseases
• Computational Biology

Background:

• Multidrug-resistant bacteria, especially Staphylococcus aureus, represent a critical global health threat.
• Accurate and timely assessment of antibiotic resistance is crucial for effective treatment and infection control.
• Early detection of resistance in S. aureus is vital to prevent transmission and guide therapeutic choices.

Purpose of the Study:

• To develop and validate a novel risk assessment framework for predicting multidrug resistance in Staphylococcus aureus isolates.
• To leverage mass spectrometry and machine learning for accurate susceptibility prediction against key antibiotics.
• To establish a multidrug resistance risk score for better clinical decision-making.

Main Methods:

• Analysis of susceptibility testing profiles for over 20,000 S. aureus isolates against six antibiotics over seven years.
• Integration of mass spectrometry data with machine learning algorithms to predict antibiotic susceptibility.
• External validation of predictive models using an independent patient cohort.

Main Results:

• High accuracy in predicting susceptibility to oxacillin, clindamycin, erythromycin, and trimethoprim-sulfamethoxazole (AUCs ranging from 0.81 to 0.94).
• Development of a multidrug resistance risk score based on predicted probabilities.
• Demonstration of the framework's utility in evaluating resistance levels and sample group performance.

Conclusions:

• The developed framework offers an efficient method for early antibiotic decision-making in S. aureus infections.
• Provides enhanced understanding of multidrug resistance risks associated with S. aureus isolates.
• Highlights the potential of integrating mass spectrometry and machine learning for clinical microbiology applications.