Multiple programmed cell death patterns predict the prognosis and therapeutic sensitivity in head and neck squamous cell carcinoma

Programmed cell death (PCD) plays a critical regulatory role in various human malignancies including head and neck squamous cell carcinoma (HNSCC); however, existing PCD-related prognostic models remain limited. This study aims to integrate multiple forms of PCD to identify more effective biomarkers for improved prognostic assessment and therapeutic response prediction in HNSCC. PCD-related genes were collected from public databases and relevant literature. Expression matrices were split into training and validation cohorts at a predetermined ratio. An autoencoder was used to extract low-dimensional features, which were then employed together with the XGBoost algorithm to construct a prognostic model. The model was validated using Kaplan–Meier survival analysis and ROC curves. Differential gene expression analysis, immune infiltration profiling, and drug sensitivity analysis were also performed. A novel multimodal programmed cell death (MPD) model was developed for HNSCC, enabling effective patient stratification into risk groups with distinct survival outcomes. The risk score was identified as an independent prognostic factor. Biologically, the low-risk group exhibited an immune-active phenotype, characterized by increased expression of immune checkpoints such as PD-1 and CTLA-4, providing a molecular basis for their potential susceptibility to immunotherapy. The model also identified differential drug sensitivity profiles, informing the design of tailored therapeutic strategies based on a patient’s MPD risk group. The MPD model helps distinguish prognosis in HNSCC patients, evaluate the benefits of immunotherapy, and supports the design of appropriate treatment strategies.