Identification of 9 Gene Signatures by WGCNA to Predict Prognosis for Colon Adenocarcinoma

BACKGROUND: A risk assessment model for prognostic prediction of colon adenocarcinoma (COAD) was established based on weighted gene co-expression network analysis (WGCNA). METHODS: From the Cancer Genome Atlas (TCGA) database, RNA-seq data and clinical data of COAD patients were retrieved. After screening of differentially expressed genes (DEGs), WGCNA was performed to identify gene modules and screen those associated with COAD progression. Then, via protein-protein interaction (PPI) network construction of module genes, hub genes were obtained, which were then subjected to the least absolute shrinkage and selection operator (LASSO) and Cox regression to build a hub gene-based prognostic scoring model. The receiver operating characteristic curve (ROC curve) was plotted for the optimal cutoff (OCO) of the risk score, based on which, patients were assigned to high or low-risk groups. Areas under the ROC curve (AUCs) were calculated, and model performance was visualized using Kaplan–Meier (KM) survival curves and verified in the external dataset GSE29621. Finally, the model's independent prognostic value was evaluated by univariate and multivariate Cox regression analyses, and a nomogram was built. RESULTS: Totally 2840 DEGs were screened from COAD dataset of TCGA, including 1401 upregulated ones and 1439 downregulated ones, which were divided into 10 modules by WGCNA. The eigenvalue of the black module was found to have a high correlation with COAD progression. PPI interaction networks were constructed for genes in the black module, and 34 hub genes were obtained by using the MCODE plug-in. A LASSO-Cox regression approach was utilized to analyze the hub genes, and a prognostic risk score model based on the signatures of 9 genes (CHEK1, DEPDC1B, FANCI, MCM10, NCAPG, PARPBP, PLK4, RAD51AP1, and RFC4) was constructed. KM analysis identified shorter overall lower survival in the high-risk group. The model was verified to have favorable predictive ability through training set and validation set. The nomogram, composed of tumor node metastasis (TNM) staging and risk score, was of good predictability. CONCLUSIONS: The COAD prognostic risk model constructed upon the signatures of 9 genes (CHEK1, DEPDC1B, FANCI, MCM10, NCAPG, PARPBP, PLK4, RAD51AP1, and RFC4) can effectively predict the survival status of COAD patients.