Detection of human saliva using surface-enhanced Raman spectroscopy combined with fractionation processing and machine learning for noninvasive screening of nasopharyngeal carcinoma

Abstract

Nasopharyngeal carcinoma (NPC) is a malignant tumor prevalent in southern China and Southeast Asia, where its early detection is crucial for improving patient prognosis and reducing mortality rates. However, existing screening methods suffer from limitations in accuracy and accessibility, hindering their application in large-scale population screening. In this work, a surface-enhanced Raman spectroscopy (SERS)-based method was established to explore the profiles of different stratified components in saliva from NPC and healthy subjects after fractionation processing. The study findings indicate that all fractionated samples exhibit disease-associated molecular signaling differences, where small-molecule (molecular weight cut-off value is 10kDa) demonstrating superior classification capabilities with sensitivity of 90.5% and specificity of 75.6%, area under receiver operating characteristic (ROC) curve of 0.925 ± 0.031. The primary objective of this study was to qualitatively explore patterns in saliva composition across groups. The proposed SERS detection strategy for fractionated saliva offers novel insights for enhancing the sensitivity and reliability of noninvasive NPC screening, laying the foundation for translational application in large-scale clinical settings.