Secondary Cross-Platform Transcriptomic Reanalysis of Laryngeal Squamous Cell Carcinoma Identifies Reproducible Tumor-Associated Programs but No Independently Validated Prognostic Biomarker
TL;DR
LSCC showed reproducible bulk-tissue expression programs, but no single differentially expressed gene qualified as an independently validated prognostic biomarker, which support biological prioritization, not clinical decision-making.
Abstract
Background: Public transcriptomes can reveal reproducible laryngeal squamous cell carcinoma (LSCC) programs, but differential expression alone does not establish prognosis. Objective: To evaluate paired tumor-associated expression changes and test whether any gene met independent prognostic criteria. Methods: Depositor-processed GSE127165 RNA-sequencing data from 57 paired LSCC and adjacent-mucosa specimens were analyzed with patient-blocked limma-trend after predefined filtering of log₂(FPKM + 1) values. Differentially expressed features required Benjamini–Hochberg false-discovery rate (FDR) <0.05 and absolute paired mean difference ≥1. Functional over-representation used g:Profiler. Expression replication used 23 paired TCGA larynx-labeled cases. Prognostic associations were assessed continuously in GSE27020 training (n=59) and validation (n=50) cohorts and 116 TCGA larynx-labeled tumors, with progression-free interval primary and overall survival secondary. Results: Of 11,908 tested features, 230 were upregulated and 205 downregulated. Excluding three quality-control-associated pairs retained 418 primary calls; quantile normalization retained 396. Among 327 measurable discovery features in TCGA, 184 replicated at FDR <0.05 with the same direction; directional concordance was 89.0% and effect correlation was ρ=0.774. Keratinization, extracellular-matrix, adhesion, and inflammatory programs were over-represented. No gene satisfied the operational prognostic criterion. SERPINE1 was exploratory: training HR=2.09, validation HR=1.53, and TCGA progression-free interval HR=1.28; the modified Knapp–Hartung pooled inference was non-significant. Conclusion: LSCC showed reproducible bulk-tissue expression programs, but no single differentially expressed gene qualified as an independently validated prognostic biomarker. These findings support biological prioritization, not clinical decision-making.