Classification of salivary based NS1 from Raman Spectroscopy with support vector machine

Area of Science:

• Biomedical Optics
• Spectroscopy
• Infectious Disease Diagnostics

Background:

• Non-Structural Protein 1 (NS1) is a key biomarker for early flavivirus infection diagnosis.
• Surface Enhanced Raman Spectroscopy (SERS) offers high sensitivity for molecular detection.
• Previous research established the SERS fingerprint of NS1 using gold substrates.

Purpose of the Study:

• To develop and validate a SERS-based method for classifying NS1-infected saliva samples from healthy controls.
• To evaluate the performance of Support Vector Machine (SVM) classifiers with different kernels for sample classification.

Main Methods:

• Analysis of saliva samples (healthy, NS1 protein, NS1-saliva mixtures) using SERS.
• Pre-processing of SERS spectra.
• Classification using SVM with linear, polynomial, and radial basis function (RBF) kernels.
• Evaluation of classifier performance based on accuracy, sensitivity, and specificity.

Main Results:

• SVM successfully classified NS1-infected and normal saliva samples.
• Polynomial and RBF kernels outperformed the linear kernel.
• The RBF kernel achieved the highest accuracy: 97.1% (100ppm), 93.4% (50ppm), and 81.5% (10ppm).

Conclusions:

• SVM classification combined with SERS is effective for distinguishing NS1-infected saliva.
• The RBF kernel demonstrates superior performance for this diagnostic application.
• This approach represents a novel method for early flavivirus infection detection.