Decoding oral leukoplakia with Raman spectroscopy: Two sample types, one signal

Oral leukoplakia (OLK) is an oral potentially malignant disorder (OPMD) with a reported global prevalence of approximately 4.1% and it is difficult to predict which lesions represent a risk. Current diagnostic pathways rely on visual inspection followed by invasive biopsy and histopathological grading, both of which are limited by sampling error, inter-observer variability, and inability to detect early biochemical changes. Raman spectroscopy has emerged as a rapid, label-free optical technique capable of detecting subtle molecular alterations associated with dysplasia and early malignant transformation. This study integrates Raman spectroscopy of exfoliated oral cells (brush biopsy) and whole saliva using both fused and multiblock analytical frameworks to improve risk stratification of OLK. Sequential orthogonalised partial least squares-discriminant analysis (SO-PLS-DA) achieved the highest accuracy, yielding sensitivities up to 97% and specificities up to 100% for distinguishing low-risk from high-risk OLK. These findings demonstrate that Raman spectroscopy provides a powerful, non-invasive diagnostic platform suitable for community-based screening, personalised surveillance, and early detection of oral cancer.