Abstract
Oral potentially malignant disorders (OPMDs) are oral mucosal conditions associated with an increased risk of oral squamous cell carcinoma (OSCC), and their clinical manifestations may be subtle and insidious. The primary aim of this study was to identify metabolite-based biomarkers for early, non-invasive detection of tumor-related metabolic signals in this at-risk population. We enrolled 21 OPMD and 46 OSCC patients, collecting saliva and plasma from all participants and tissue samples from a subset of the same cohort (12 OPMD and 5 OSCC). Untargeted metabolomics identified 491 and 303 differential metabolites in saliva and plasma, respectively. Five metabolites-dodecanoic acid, tetradecanedioic acid, porphobilinogen, uridine, and isocitrate-were significantly altered in both biofluids. Tissue validation showed significant alterations in dodecanoic acid, tetradecanedioic acid, and isocitrate. Using all biologically annotated differential metabolites, AUCs were 0.888 (saliva) and 0.994 (plasma), while the tissue-anchored three-metabolite classifier yielded 0.694 (saliva) and 0.852 (plasma). Full-length 16S rDNA sequencing and integrative microbiome-metabolome analysis indicated potential correlations between microbial shifts and metabolite profiles. Our findings highlight metabolic alterations and cross-compartment associations across saliva, plasma, and tissue in OPMDs and OSCC. Notably, despite the higher internal discrimination of all-differential models, the tumor-informed three-metabolite model captures tissue-aligned metabolic changes detectable in saliva and plasma, providing a specific, non-invasive readout for early detection.