Targeted Surface-Enhanced Raman Scattering for Highly Accurate Identification of Bacterial Species and Finding...

Area of Science:

• Biophotonics
• Spectroscopy
• Artificial Intelligence

Background:

• Accurate bacterial identification is crucial for sepsis prevention.
• Conventional methods face limitations in speed and specificity.
• Raman spectroscopy, enhanced by AI, offers a promising alternative.

Purpose of the Study:

• To develop and optimize a SERS-based method for rapid and accurate bacterial identification.
• To investigate the influence of nanoparticle properties and excitation wavelength on SERS spectra.
• To establish a framework for interpreting AI algorithms in SERS analysis.

Main Methods:

• Utilized colloidal gold (Au) and silver (Ag) nanoparticles (NPs) for reproducible SERS spectra acquisition.
• Analyzed spectra within the biological fingerprint region (500-1300 cm^-1).
• Employed a deep neural network model with mannose-modified AuNPs and 532 nm excitation for classification.
• Applied normalized positive Shapley additive explanations (npSHAP) for AI interpretation.

Main Results:

• Achieved 96.1% classification accuracy for 14 bacterial species using mannose-modified AuNPs.
• Demonstrated the effectiveness of targeted SERS with specific nanoparticle modifications.
• Identified key spectral features (top five peaks) as an AI-selected spectral barcode.

Conclusions:

• The proposed SERS approach with AI interpretation provides a robust platform for bacterial identification.
• The developed framework enhances the interpretability of AI in spectral analysis.
• This method holds potential for improving clinical diagnostics and sepsis management.