Machine Learning Modelling, Single-Cell Landscape Profiling and Spatial Transcriptomics Provide New Insights Into SUMOylation in Head and Neck Squamous Cell Carcinoma

Abstract

SUMOylation is implicated in the regulation of multiple malignancies. However, its potential roles in head and neck squamous cell carcinoma (HNSCC) remain insufficiently characterised. By integrating bulk RNA-seq, scRNA-seq and stRNA-seq datasets, we systematically interrogated the biological relevance of SUMOylation in HNSCC. Key markers from the signature were further validated using in vitro functional assays. A recognition model was established and validated using 692 HNSCC and 178 non-HNSCC samples. SROC analysis demonstrated robust performance across eight datasets (AUC = 0.92). Functional enrichment and scRNA-seq analyses indicated that SUMOylation may exert its effects in HNSCC primarily through cell-cycle regulation. Among the model features, SAE1 was markedly overexpressed in HNSCC (SMD = 1.07, 95% CI 0.56–1.59, p < 0.05). In vitro assays further confirmed that SAE1 enhanced proliferation, colony formation and migration in SAS and SCC-9 cells and validated its role in regulating cell cycle progression and apoptosis. We established a SUMOylation-related recognition model for HNSCC with consistently strong performance in multiple external validation cohorts. SAE1 emerges as a candidate molecular biomarker for HNSCC.