Integrative single-cell and machine learning analysis identifies a tumor doubling time-related prognostic signature and therapeutic targets in head and neck squamous cell carcinoma.
Journal:
Translational oncology
Published Date:
Jul 23, 2026
Abstract
BACKGROUND: Head and neck squamous cell carcinoma (HNSCC) exhibits marked molecular heterogeneity and diverse clinical outcomes, partly influenced by HPV status and the tumor immune microenvironment. Although tumor doubling time (TDT) reflects tumor growth kinetics and has prognostic relevance, its molecular basis in HNSCC remains unclear. METHODS: Single-cell RNA-seq data from GSE164690 were integrated with bulk transcriptomic datasets from TCGA-HNSC, GSE41613, GSE65858, and ICGC. We identified 21 tumor doubling time-related genes (TDTRGs), evaluated their cell type-specific enrichment using UCell, and defined molecular subtypes by consensus clustering. Immune infiltration was assessed by ssGSEA and CIBERSORT. A prognostic model was developed using 117 machine-learning algorithm combinations in the Mime1 package and further evaluated by single-cell visualization, cell cycle scoring, pseudotime analysis, drug sensitivity prediction, and molecular docking. RESULTS: A single-cell atlas comprising 17 cell populations was constructed. TDTRGs were mainly enriched in cycling T cells and epithelial cells and showed HPV-related expression differences. Two molecular subtypes were identified; C2 was associated with poorer progression-free interval (P = 0.001) and enrichment of MET and integrin signaling. A seven-gene signature-CCBE1, MAD2L1, ECT2, BIRC5, ITGA6, ESM1, and NUF2-showed robust prognostic performance across independent cohorts and outperformed 73 published signatures. Single-cell and pseudotime analyses confirmed cell type-specific and dynamic regulation. Venetoclax was predicted as a potential therapeutic candidate, with docking energies below -9 kcal·mol⁻¹. CONCLUSION: This study defines TDTRG-related heterogeneity in HNSCC and establishes a robust seven-gene prognostic signature, providing potential biomarkers and therapeutic insights for personalized management.
Authors
Keywords
No keywords available for this article.