Multimodal Integration Identifies C16orf74 as an EMT Regulator With Prognostic Relevance in HNSCC

Head and neck squamous cell carcinoma (HNSCC) is an aggressive malignancy characterized by marked intratumoral heterogeneity and poor clinical outcomes. Epithelial-mesenchymal transition (EMT) promotes metastasis, immune evasion, and therapeutic resistance. However, the key molecular determinants underlying EMT-associated malignant states remain incompletely defined, and their translation into clinically accessible tools is challenging. In this study, we performed an integrative multiomics analysis to dissect malignant epithelial heterogeneity in HNSCC. Single-cell RNA sequencing (scRNA-seq) identified a highly malignant EMT epithelial subpopulation characterized by elevated genomic instability and terminal malignant states. Furthermore, an ensemble of machine learning algorithms prioritized C16orf74 as a key EMT-associated malignant regulator. Spatial transcriptomics localized C16orf74 spatially concentrated at invasive-front and transitory regions associated with distinct tumor microenvironmental niches. Gain- and loss-of-function assays confirmed that C16orf74 regulates EMT activation, migration, invasion, and tumor-sphere formation in HNSCC cells. Mechanistically, integrative regulon, transcriptomic, spatial, and functional analyses identified high-mobility group AT-hook 2 (HMGA2) as a candidate downstream mediator of C16orf74-associated EMT. To translate these biological insights into clinical practice, we developed an EMT-based pathomics prognostic model using hematoxylin and eosin-stained whole-slide images, which achieved robust prognostic stratification. Notably, high-risk group interpretably exhibited a positive correlation with C16orf74 expression, EMT enrichment, an immunosuppressive microenvironment, and resistance to standard chemotherapeutic agents. Collectively, our findings identify C16orf74 as a putative EMT regulator and develop a clinically accessible pathomics model for prognostic and therapeutic stratification in HNSCC.