BACKGROUND

Recently, endoplasmic reticulum stress related gene (ERS) markers have performed very well in predicting the prognosis of tumor patients.

METHODS

The differentially expressed genes in Oral squamous cell carcinoma (OSCC) were obtained from TCGA and GTEx database. Three prognosis-related and differentially expressed ERSs were screened out by Least Absolute Selection and Shrinkage Operator (Lasso) regression to construct a prognostic risk model. Receiver Operating Characteristic Curve (ROC), riskplots and survival curves were used to verify the model's accuracy in predicting prognosis. Multi-omics analysis of immune infiltration, gene mutation, and stem cell characteristics were performed to explore the possible mechanism of OSCC. Finally, we discussed the model's clinical application value from the perspective of drug sensitivity.

RESULTS

Three genes used in the model (IBSP, RDM1, RBP4) were identified as prognostic risk factors. Bioinformatics analysis, tissue and cell experiments have fully verified the abnormal expression of these three genes in OSCC. Multiple validation methods and internal and external datasets confirmed the model's excellent performance in predicting and discriminating prognosis. Cox regression analysis identified risk score as an independent predictor of prognosis. Multi-omics analysis found strong correlations between risk scores and immune cells, cell stemness index, and tumor mutational burden (TMB). It was also observed that the risk score was closely related to the half maximal inhibitory concentration of docetaxel, gefitinib and erlotinib. The excellent performance of the nomogram has been verified by various means.

CONCLUSION

A prognostic model with high clinical application value was constructed. Immune cells, cellular stemness, and TMB may be involved in the progression of OSCC.