Establishment of Chemotherapy Prediction Model Based on Hypoxia-Related Genes for Oral Cancer.

Purpose: Identify the hypoxia genes related to chemotherapy resistance of oral cancer, and construct a chemotherapy response model by machine learning algorithm. Methods: 72 oral cancer patients with complete chemotherapy records and chemotherapy reactions were screened from the Cancer Genome Atlas (TCGA) database. According to the chemotherapy reactions, they were divided into chemotherapy sensitive group and chemotherapy resistant group. The differential genes were screened by Limma package. Then the chemotherapy response gene were screened by univariate analysis. Based on the gene expression profile of chemotherapy response, four machine learning algorithms were used to construct the prediction model of chemotherapy response. The core genes were screened by lasso regression analysis. Finally, the prognosis and immune infiltration of the core genes were analyzed. The results were verified by immunohistochemistry (IHC). Results: We obtained 22 hypoxia related differential genes. Univariate analysis found 6 Chemotherapy response genes. Machine learning algorithms show that XGBoost have the best predictive performance for chemotherapy response. ALDOA is the core gene of chemotherapy resistance. Conclusions: Successfully constructed a chemotherapy prediction model for oral cancer by machine learning algorithm. Under hypoxia, the high expression of ALDOA is associated with chemotherapy resistance in oral cancer.