Head and neck squamous cell carcinoma (HNSCC) is a highly aggressive cancer with poor prognosis. Although NOTCH signaling is frequently inactivated in HNSCC, the prognostic significance of NOTCH-related genes and their link to the tumor microenvironment remain unclear.

NOTCH-related genes were identified in HNSCC patients through univariate Cox regression and differential expression analyses. A prognostic risk model (NOTCH score) was constructed using LASSO regression. The association between the NOTCH score and clinical outcomes, immune infiltration, therapy response sensitivity was comprehensively evaluated. Single-cell RNA sequencing and immunohistochemistry were used to assess the cellular and tissue distribution of risk genes.

Seven candidate genes (COL28A1, EFEMP1, FGF9, FMOD, ITGA5, ROBO1, and THBS1) were selected to construct the NOTCH-based prognostic model. Patients were stratified into high- and low-risk groups based on the median NOTCH score. High-risk patients exhibited significantly worse survival compared to the low-risk group. ROC analysis demonstrated robust prognostic performance, with AUC values exceeding 0.7 for 3-year survival in both training and validation cohorts. High NOTCH scores were associated with an immunosuppressive tumor microenvironment and poor response to immunotherapy. Drug sensitivity analysis revealed that high-risk patients were more sensitive to agents such as CMK, Parthenolide, and Thapsigargin. At both the single-cell and protein levels, the seven genes were predominantly expressed in fibroblasts and endothelial cells, suggesting their potential involvement in tumor stromal remodeling and microenvironment regulation.

We developed and validated a NOTCH-based prognostic model with strong predictive power in HNSCC. The NOTCH score may help assess immune status, predict treatment response, and guide personalized therapy.