A multi-omics pipeline integrating machine learning and spatial-cellular analysis identifies SASH1 as a prognostic biomarker and therapeutic target in head and neck squamous cell carcinoma.

Head and neck squamous cell carcinoma (HNSCC) is a highly aggressive malignancy with a poor prognosis, necessitating the discovery of novel and reliable molecular biomarkers for improved clinical management. Traditional bulk transcriptomic analyses often mask the cellular heterogeneity and spatial complexity of the tumor microenvironment, limiting the identification of robust biomarkers. This study aimed to identify and validate key driver genes in HNSCC through a comprehensive multi-omics and machine learning-based approach. Transcriptomic data from multiple GEO datasets (GSE29330, GSE6631, GSE138206) and the TCGA-HNSC cohort were integrated and analyzed to identify consensus differentially expressed genes (DEGs). A suite of four machine learning algorithms (LASSO, SVM-RFE, XGBoost, Boruta) was employed to screen for core candidate genes. The cellular origins and spatial distribution of these core genes were subsequently dissected using public single-cell (GSE215403) and spatial transcriptomics (GSE252265) data. Finally, the expression of the key gene, SAM and SH3 domain-containing 1 (SASH1), was validated at the protein level via Western blot in HNSCC cell lines, and its clinical and therapeutic value was assessed through survival, clinical correlation, and drug sensitivity analyses. An integrated analysis of bulk transcriptomic data identified 159 consensus DEGs, from which four core genes (COL1A1, EMP1, MYH11, SASH1) were robustly selected by all four machine learning algorithms. Multi-omics validation revealed that SASH1 was specifically downregulated within the malignant cell population and its expression was spatially exclusive from the COL1A1-high fibrotic stromal regions. Western blot confirmed the significant downregulation of SASH1 protein in HNSCC cells compared to controls. Importantly, low SASH1 expression was significantly associated with poorer overall survival in the TCGA cohort ( P < 0.05), a prognostic value not observed for the other core genes. Functional analyses linked SASH1 to critical pathways including cell cycle and adhesion. Furthermore, SASH1 expression levels correlated with sensitivity to multiple targeted drugs, including ATR and Aurora kinase inhibitors. By systematically integrating multi-platform transcriptomics, machine learning, and multi-dimensional validation, this study identifies SASH1 as a robust prognostic biomarker and a potential predictor of therapeutic response in HNSCC. The established multi-omics pipeline provides a meaningful framework for biomarker discovery and highlights SASH1 as a promising target for advancing precision medicine in HNSCC.