PCA-QDA model selection for detecting NS1 related diseases from SERS spectra of salivary mixtures

Abstract

Of recent, non-structural protein (NS1) in saliva has emerged to be engaging as a detection biomarker for diseases related to NS1 at febrile stage. Non-invasive detection of NS1 in saliva, free from risk of blood infection, further will make the approach more preferred than the current serum based ones. Our work here intends to define an optimal classifier model for Quadratic Discriminant Analysis (QDA), optimized with Principal Component Analysis (PCA), to distinct between positive and negative NS1 adulterated samples from salivary SERS spectra. The adulterated samples are acquired from our UiTM-NMRR-12-1278-12868-NS1-DENV database. Then, PCA extracts significant features from the database after pre-processing, based on three stopping criteria, which are served as inputs to the QDA classifiers. It is found that the PCA-QDA pseudo model with 5, 70 and 115 principal components from the three criterion achieves performance of 100% (Scree), 84.2% (CPV) and 55.3% (EOC) in accuracy. Higher accuracy at 100% (Scree), 97.3684% (CPV) and 97.3684% (EOC) are observed with QDA diagonal model.