Machine learning-based prognostic modeling integrating PANoptosis in head and neck squamous cell carcinoma

Head and neck squamous cell carcinoma (HNSCC) is a highly heterogeneous and aggressive cancer, posing challenges for prognosis and treatment. The role of PANoptosis in HNSCC remains unclear, despite its potential impact on the tumor immune microenvironment.

We obtained the single-cell RNA sequencing dataset GSE181919 from the GEO database and used the AUCell R package to analyze PANoptosis enrichment and heterogeneity, identifying key PANoptosis-related genes at the single-cell level. A prognostic model was constructed using 101 machine learning algorithms and their combinations, with TCGA-HNSCC as the training set and GSE41613 and GSE65858 as validation sets. Model performance was assessed using Kaplan-Meier survival analysis, ROC curves, and PCA. Glutathione S-transferase omega 1 (GSTO1) was identified as a key model gene, and its expression was validated using PCR in clinical samples.

We integrated single-cell sequencing, bulk transcriptomic sequencing, and machine learning algorithms to construct a StepCox[backward] + RSF prognostic model for HNSCC. This model stratified HNSCC patients into high-risk and low-risk groups, with the high-risk group exhibiting worse prognosis. ROC and PCA analyses confirmed the model's robustness. Additionally, the key gene GSTO1 was identified and further validated to be upregulated in HNSCC and associated with poor prognosis.

The PANoptosis-based prognostic model offers strong predictive value for HNSCC and has potential applications in personalized treatment approaches. GSTO1 stands out as a promising biomarker and therapeutic target. Future research should focus on experimental validation and the development of therapeutic strategies that modulate PANoptosis to improve outcomes for HNSCC patients.