Machine Learning-Driven Surface-Enhanced Raman Spectroscopy (SERS) Profiling of Low Molecular Weight Serum from Cervical Cancer Patients

Cervical cancer is a leading contributor to both mortality and morbidity in low- and middle-income countries with Human papillomavirus (HPV) which is a small, nonenveloped virus containing a double-stranded circular DNA genome, identified as the predominant cause. The survival rate increases up to 95% when cervical cancer is detected through early-stage screening. In this work, surface-enhanced Raman spectroscopy (SERS), an enhanced version of Raman spectroscopy, using silver nanoparticles (Ag-NPs) as a SERS substrate, has been explored for the rapid monitoring of biochemical variations identified among the various stages of cervical carcinoma patients as compared to healthy ones. Human blood is a rich source of biomolecules, in which high-molecular-weight fractions (HMWF) often mask low-molecular-weight fractions (LMWF), which contain disease-related biomarkers. Centrifugal filtration with a 100 kDa cutoff value is performed to separate HMWF from LMWF for SERS spectral identification and analysis of disease biomarkers. The SERS spectral features of 100 kDa filtrate fractions of disease samples are found at 353, 562, 766, 1180, 1307, 1365, 1508, and 1653 cm−1, as compared to peaks visible in healthy filtrate samples. Furthermore, the SERS spectra of various samples (cancerous and controlled) have been classified by principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA). Using validated PLS-DA analysis, the normal and cancerous samples are successfully identified and distinguished with 99% sensitivity and 95% specificity, and the K-Nearest Neighbor (KNN) model is used to check similarity between samples which confirms the mean accuracy of 94.79%.