Gene and Isoform Misassignment Among Candidate TP53 and CDKN2A Variants of Uncertain Significance in Oral Squamous Cell Carcinoma: A Public-Database Curation Study

Background: Variants of uncertain significance (VUSs) in TP53 and CDKN2A are reported often in oral squamous cell carcinoma (OSCC), yet variant-level evidence tends to lag behind established gene-level associations. Candidate variant lists compiled by automated literature synthesis, registry mining, or AI-assisted extraction carry a risk of gene-level or transcript-level mis-annotation that is not widely recognized. Methods: Ten candidate VUSs (one TP53 and nine CDKN2A), drawn from a candidate list generated by an AI-assisted literature and registry search (SciSpace, PubMed, Google Scholar, and ClinicalTrials.gov), were curated using ClinVar, dbSNP/NCBI Gene, gnomAD (via VarSome), and UCSC phyloP conservation tracks. To address regional population representation, two Indian genomic resources (dbGENVOC and TMC-SNPdb 2.0) and one OSCC-specific catalogue (dbGVOSCC) were also queried. Gene and transcript assignment was verified independently for every variant before evaluation against American College of Medical Genetics and Genomics and Association for Molecular Pathology (ACMG/AMP) criteria. Results: The candidate TP53 variant (rs2543751607) is a coding variant of the adjacent gene WRAP53 and is non-coding relative to TP53; it was excluded. All CDKN2A variants mapped only to the p14(ARF) transcript (NM_058195.4) and not to the p16(INK4a) transcript (NM_000077.5) that underlies the CDK4/6-inhibitor predictive-biomarker role of CDKN2A. One rsID (rs2489327173) resolved to two distinct missense variants, giving ten evaluated CDKN2A variants: two Likely Pathogenic (provisional, isoform-qualified), seven Uncertain Significance, and one Uncertain Significance and leaning benign. None of the ten was recorded in dbGENVOC, TMC-SNPdb 2.0, or dbGVOSCC. In dbGENVOC the only CDKN2A coding variants in Indian OSCC patients fell in the shared exon 2, and in dbGVOSCC the curated OSCC CDKN2A events were mainly deletions and p16 methylation. Curation also corrected an allele-matching error at a multiallelic position and resolved two conservation-score discordances by independent phyloP querying. Conclusions: Public-database curation of a small candidate VUS set exposed structural annotation problems, namely gene misassignment and isoform misassignment, that change the clinical interpretation of the source synthesis. Variant lists derived automatically or from the literature warrant independent gene-level and transcript-level verification, and CDKN2A isoform specificity should be stated explicitly whenever it is discussed as a biomarker in OSCC or HNSCC.