FTIR-Derived Feature Insights for Predicting Time-Dependent Antibiotic Resistance Progression

Area of Science:

• Microbiology
• Spectroscopy
• Computational Biology

Background:

• Antibiotic resistance is a growing global health threat.
• Conventional diagnostics often treat resistance as a binary trait, failing to capture its dynamic nature.
• Understanding the temporal progression of resistance is crucial for effective treatment.

Purpose of the Study:

• To investigate the utility of Fourier-transform infrared spectroscopy (FTIR) and machine learning (ML) for monitoring the dynamic development of antibiotic resistance.
• To determine how spectral features and analysis parameters influence the accuracy of resistance classification.
• To explore the potential for real-time monitoring of bacterial adaptation to antibiotics.

Main Methods:

• Fourier-transform infrared spectroscopy (FTIR) was used to collect spectral data from *Staphylococcus aureus*.
• Bacteria were exposed to azithromycin, trimethoprim/sulfamethoxazole, and oxacillin over time during resistance induction.
• Principal component analysis (PCA) and machine learning algorithms (ML) were applied to FTIR spectra to identify resistance patterns.

Main Results:

• FTIR-based biochemical profiling, integrated with ML, accurately identified antibiotic resistance.
• Classification accuracy reached up to 96%, depending on the antibiotic and analytical parameters.
• Early signs of resistance were detected as early as 24 hours post-exposure.
• Combining multiple spectral regions significantly improved model performance compared to single regions.

Conclusions:

• Antibiotic resistance should be considered a dynamic, time-dependent adaptive trajectory.
• FTIR coupled with ML offers a powerful tool for real-time monitoring of resistance.
• This approach supports adaptive antimicrobial management and personalized therapeutic decisions through spectral biomarkers.