Systems toxicology and multi-omics prioritize an MTBC-related computational signature in oral squamous cell carcinoma

Computational prioritization can nominate candidate constituents from complex exposures, but it does not establish exposure or causality. We integrated an areca-constituent library with archived ADMET predictions and oral squamous cell carcinoma (OSCC) transcriptomic data to evaluate 2-methyl-1,2,3,4-tetrahydro-beta-carboline (MTBC) as a candidate for further study. Replacing GTEx salivary controls with 44 TCGA-HNSC adjacent normals reduced the OSCC Judge Gene Set from 2,108 to 773 genes and yielded an 11-gene MTBC-related computational signature. In bulk data, this signature was not more tumour-discriminatory than expression-matched random Judge Gene Sets. Patient-level single-cell analysis did not support the previously reported positive association with epithelial state, whereas a cross-tissue cutaneous SCC analysis showed spatial alignment with TLS scores but did not validate an oral or areca-specific effect. A leakage-safe penalized classifier distinguished 51 oral tumours from 44 adjacent normals in internal evaluation (development nested-CV AUC 0.964, 95% CI 0.916-0.997), but no external or screening claim is made. These results define a reproducible computational prioritization framework and an exploratory MTBC-related expression signature that requires direct exposure measurement and experimental validation.