Objective To screen metabolic core genes in colon cancer based on machine learning algorithms and analyze their functional mechanisms.

Methods Data were obtained from The Cancer Genome Atlas (TCGA) database and the Gene Expression Omnibus (GEO) database. The TCGA cohort included 375 tumor samples and 32 adjacent normal tissue samples, while the GSE39582 cohort comprised 419 tumor samples. Univariate Cox regression analysis combined with random forest, support vector machine recursive feature elimination (SVM-RFE), and least absolute shrinkage and selection operator (LASSO) regression algorithms were employed to screen for metabolic core genes. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was used to evaluate the predictive efficacy of the core genes. Real-time fluorescent quantitative polymerase chain reaction (qRT-PCR) and immunohistochemistry (IHC) methods were adopted to detect the expression of the core genes. The core genes were knocked down to explore their roles in colon cancer.

Results Three core genes, namely CPT2, SCP2 and NR3C2 , were screened based on machine learning algorithms. According to the comparison results of the AUCs of the ROC curves, NR3C2 exhibited the best predictive efficacy. qRT-PCR detection results showed that NR3C2 mRNA was lowly expressed in colon cancer cell lines; IHC detection results revealed that NR3C2 was lowly expressed in colon cancer tissues. Knocking down NR3C2 significantly promoted the proliferation and migration of colon cancer cells.

Conclusion NR3C2 is identified as a core metabolic inhibitory gene in colon cancer by cross-applying three machine learning algorithms, which may provide a new strategy for metabolic targeted therapy.