A tricarboxylic acid cycle-based machine learning model to select effective drug targets for the treatment of esophageal squamous cell carcinoma

Background: The tricarboxylic acid cycle (TCA cycle) is an important metabolic pathway and closely related to tumor development. However, its role in the development of esophageal squamous cell carcinoma (ESCC) has not been fully investigated. Methods: The RNA expression profiles of ESCC samples were retrieved from the TCGA database, and the GSE53624 dataset was additionally downloaded from the GEO database as the validation cohort. Furthermore, the single cell sequencing dataset GSE160269 was downloaded. TCA cycle-related genes were obtained from the MSigDB database. A risk score model for ESCC based on the key genes of the TCA cycle was built, and its predictive performance was evaluated. The association of the model with immune infiltration and chemoresistance were analyzed using the TIMER database, the R package “oncoPredict” score, TIDE score and so on. Finally, the role of the key gene CTTN was validated through gene knockdown and functional assays. Results: A total of 38 clusters of 8 cell types were identified using the single-cell sequencing data. The cells were divided into two groups according to the TCA cycle score, and 617 genes were identified that were most likely to influence the TCA cycle. By intersecting 976 key genes of the TCA cycle with the results of WGCNA, 57 genes significantly associated with the TCA cycle were further identified, of which 8 were screened through Cox regression and Lasso regression to construct the risk score model. The risk score was a good predictor of prognosis across subgroups of age, N, M classification and TNM stage. Furthermore, BI-2536, camptothecin and NU7441 were identified as possible drug candidates in the high-risk group. The high-risk score was associated with decreased immune infiltration in ESCC, and the low-risk group had better immunogenicity. In addition, we also evaluated the relationship between risk scores and immunotherapy response rates. Functional assays showed that CTTN may affect the proliferation and invasion of ESCC cells through the EMT pathway. Conclusion: We constructed a predictive model for ESCC based on TCA cycle-associated genes, which achieved good prognostic stratification. The model are likely associated with the regulation of tumor immunity in ESCC.