Background: Head and neck squamous cell carcinoma (HNSCC) is a malignant tumor with a very high mortality rate, and a large number of studies have confirmed the correlation between inflammation and malignant tumors and the involvement of inflammation-related regulators in the progression of HNSCC. However, a prognostic model for HNSCC based on genes involved in inflammatory factors has not been established. Methods: First, we downloaded transcriptome data and clinical information from patients with head and neck squamous cell carcinoma from TCGA and GEO (GSE41613) for data analysis, model construction, and differential gene expression analysis, respectively. Genes associated with inflammatory factors were screened from published papers and intersected with differentially expressed genes to identify differentially expressed inflammatory factor-related genes. Subgroups were then typed according to differentially expressed inflammatory factor-related genes. Univariate, LASSO and multivariate Cox regression algorithms were subsequently applied to identify prognostic genes associated with inflammatory factors and to construct prognostic prediction models. The predictive performance of the model was evaluated by Kaplan-Meier survival analysis and receiver operating characteristic curve (ROC). Subsequently, we analyzed differences in immune composition between patients in the high and low risk groups by immune infiltration. The correlation between model genes and drug sensitivity (GSDC and CTRP) was also analyzed based on the GSCALite database. Finally, we examined the expression of prognostic genes in pathological tissues, verifying that these genes can be used to predict prognosis. Results: Using univariate, LASSO, and multivariate cox regression analyses, we developed a prognostic risk model for HNSCC based on 13 genes associated with inflammatory factors (ITGA5, OLR1, CCL5, CXCL8, IL1A, SLC7A2, SCN1B, RGS16, TNFRSF9, PDE4B, NPFFR2, OSM, ROS1). Overall survival (OS) of HNSCC patients in the low-risk group was significantly better than that in the high-risk group in both the training and validation sets. By clustering, we identified three molecular subtypes of HNSCC carcinoma (C1, C2, and C3), with C1 subtype having significantly better OS than C2 and C3 subtypes. ROC analysis suggests that our model has precise predictive power for patients with HNSCC. Enrichment analysis showed that the high-risk and low-risk groups showed strong immune function differences. CIBERSORT immune infiltration score showed that 25 related and differentially expressed inflammatory factor genes were all associated with immune function. As the risk score increases, specific immune function activation decreases in tumor tissue, which is associated with poor prognosis. We also screened for susceptibility between the high-risk and low-risk groups and showed that patients in the high-risk group were more sensitive to talazoparib-1259, camptothecin-1003, vincristine-1818, Azd5991-1720, Teniposide-1809, and Nutlin-3a (-) -1047.Finally, we examined the expression of OLR1, SCN1B, and PDE4B genes in HNSCC pathological tissues and validated that these genes could be used to predict the prognosis of HNSCC. Conclusion: In this experiment, we propose a prognostic model for HNSCC based on inflammation-related factors. It is a non-invasive genomic characterization prediction method that has shown satisfactory and effective performance in predicting patient survival outcomes and treatment response. More interdisciplinary areas combining medicine and electronics will be explored in the future.
Read full abstract