Machine learning enabled multiplex detection of periodontal pathogens by surface-enhanced Raman spectroscopy

Area of Science:

• Microbiology
• Nanotechnology
• Spectroscopy

Background:

• Periodontitis is a chronic inflammatory disease driven by bacterial infections, leading to tooth-supporting structure destruction.
• Salivary microbial analysis is crucial for assessing periodontal health.
• Existing bacterial detection methods are often complex and require labeling.

Purpose of the Study:

• To develop a rapid, label-free, and multiplexed detection method for key periodontal pathogens.
• To characterize the nanoscale dimensions of periodontal pathogens using atomic force microscopy.
• To establish a machine learning-based detection system using surface-enhanced Raman spectroscopy.

Main Methods:

• Atomic force microscopy (AFM) was used to determine the nanoscale dimensions of Actinobacillus actinomycetemcomitans, Porphyromonas gingivalis, and Streptococcus mutans.
• Surface-enhanced Raman spectroscopy (SERS) was employed to obtain unique, label-free Raman signatures for each bacterial species.
• Machine learning models were trained on SERS spectra for accurate identification and quantification of bacteria in pure and mixed samples.

Main Results:

• Periodontal pathogens were characterized as nanorods with dimensions of 0.6-1.1 μm in length and 500-700 nm in width.
• Distinct SERS spectra were identified for each target pathogen, enabling label-free detection.
• A machine learning model achieved 95.6% accuracy in simultaneously identifying all three tested pathogens in mixed samples.

Conclusions:

• SERS combined with machine learning offers a highly accurate and efficient platform for label-free, multiplexed detection of periodontal pathogens.
• This method significantly simplifies sample preparation and speeds up bacterial identification in saliva.
• The developed sensing modality holds promise for widespread clinical application in diagnosing and monitoring periodontitis.