Clinical Human-Derived Pathogen Signatures Captured by SERS and Deep Learning for Environmental Exposure Risk...

Area of Science:

• Environmental Science
• Microbiology
• Spectroscopy

Background:

• Microbial contamination and antibiotic-resistant bacteria (ARB) are significant environmental threats, especially in areas with poor sanitation.
• Urinary pathogens in contaminated environments serve as early indicators of ARB exposure.
• Existing exposure assessment methods using standard strains are limited for real-world applications.

Purpose of the Study:

• To develop and validate a label-free platform for ARB exposure assessment using SERS and CNN.
• To create a comprehensive spectral database of clinical urinary isolates for training and testing the model.
• To enhance the interpretability of the ARB detection model.

Main Methods:

• Integration of surface-enhanced Raman spectroscopy (SERS) with a convolutional neural network (CNN).
• Development of a spectral database from 368 clinical urinary isolates.
• Validation of the CNN model on wastewater and independent urine samples.
• Utilizing SHAP for model interpretation to identify discriminatory spectral features.

Main Results:

• The CNN model achieved a high classification accuracy of 97.6% on clinical isolates.
• The platform demonstrated robust performance in wastewater (92.2%) and urine samples (90.3%).
• SHAP analysis identified spectral features linked to adenine and tryptophan, improving model transparency.

Conclusions:

• A practical and transparent framework for microbial surveillance and ARB exposure assessment was established.
• The SERS-CNN platform offers real-time monitoring capabilities for environmental health management.
• This approach provides a valuable tool for detecting ARB in complex environmental matrices.