Construction and validation of a prognostic model for colon adenocarcinoma based on bile acid metabolism-related genes

Abstract

AbstractColon adenocarcinoma (COAD), one of the common clinical cancers, exhibits high morbidity and mortality, and its pathogenesis and treatment are still underdeveloped. Numerous studies have demonstrated the involvement of bile acids in tumour development, while the potential role of their metabolism in the tumor microenvironment (TME) has not been explored. A collection of 481 genes related to bile acid metabolism were obtained, and The Cancer Genome Atlas-based COAD risk model was developed using the least absolute shrinkage selection operator (LASSO) regression analysis. The Gene Expression Omnibus dataset was used to validate the results. The predictive performance of the model was verified using column line plots, principal component analysis and receiver operating characteristic curves. Then, we analysed the differences between the high- and low-risk groups from training set based on clinical characteristics, immune cell infiltration, immune-related functions, chemotherapeutic drug sensitivity and immunotherapy efficacy. Additionally, we constructed a protein–protein interaction network to screen for target genes, which were further investigated in terms of differential immune cell distribution. A total of 234 bile acids-related differentially expressed genes were obtained between normal and tumour colon tissues. Among them, 111 genes were upregulated and 123 genes were down-regulated in the tumour samples. Relying on the LASSO logistic regression algorithm, we constructed a model of bile acid risk score, comprising 12 genes: CPT2, SLCO1A2, CD36, ACOX1, CDKN2A, HADH, GABRD, LEP, TIMP1, MAT1A, SLC6A15 and PPARGC1A. This model was validated in the GEO-COAD set. Age and risk score were observed to be independent prognostic factors in patients with COAD. Genes related to bile acid metabolism in COAD were closely related to bile secretion, intestinal transport, steroid and fatty acid metabolism. Furthermore, the high-risk group was more sensitive to Oxaliplatin than the low-risk group. Finally, the three target genes screened were closely associated with immune cells. We identified a set of 12 genes (CPT2, SLCO1A2, CD36, ACOX1, CDKN2A, HADH, GABRD, LEP, TIMP1, MAT1A, SLC6A15, and PPARGC1A) associated with bile acid metabolism and developed a bile acid risk score model using LASSO regression analysis. The model demonstrated good predictive performance and was validated using an independent dataset. Our findings revealed that the bile acid risk score were independent prognostic factors in COAD patients.