In the present study, 5 significantly stable gene modules related to colon cancer were constructed by WGCNA algorithm. Then, 1153 consistently DEGs were identified between colon cancer tumor and normal tissues samples based on the TCGA, GSE44861 and GSE44076 datasets. Furthermore, based on the intersected genes between genes of gene module by WGCNA and consistently DEGs by MetaDE, 12 DEGs (ADORA3, CPA3, CPM, EDN3, FCRL2, MFNG, NAT1, PCSK5, PPARGC1A, PRRX2, TNFRSF17, and WDR78) related to prognosis of colon cancer were further isolated as the optimized prognostic gene signature, and a prognostic model was constructed based these 12 DEGs, which presented a relative highly forecast ability for the prognosis of colon cancer both in the training dataset and validation datasets. In addition, age, pathological T, tumor recurrence, and prognostic model status were identified as the independent prognostic factors in patients with colon cancer based on TCGA. Furthermore, based on the prognostic model, 514 DEGs related to prognosis of colon cancer were further identified, which were closely associated with ion transport, cell-cell signaling, regulation of cyclic nucleotide metabolic process, neuroactive ligand-receptor interaction, and calcium signaling pathway.
The mining of a large amount of genetic data in various diseases have been enhanced due to the rapid technological advances in high-throughput sequencing and bioinformatics (23). TCGA, as a public and available cancer genomic datasets, provides the comprehensive data of cancers, including mRNA expression data, miRNA expression data, copy number variation, DNA methylation, and clinical information (24). The data from TCGA have been effectively applied to improve diagnostic and therapeutic methods of cancers, as well as finally cancer prevention (24). Thus, this study was performed based on the gene expression profile data and clinical information of BC form TCGA and GEO database. Gene expression profiles have been reported to predict the prognosis outcome of cancers (25–27). Computationally, univariate and multivariate Cox regression were the most common method to construct the prognostic models and screen prognostic factors (28). In this study, the Cox regression model based on the LASSO, a semi-parametric proportional hazards model, was applied. The availability of this model in survival analysis have been confirmed in recent studies (29, 30). Similarly, in this study, the prognostic model constructed by LASSO Cox regression model showed a higher predictive ability both in training and validation sets. In addition, this study showed that age pathological T, and tumor recurrence were independent prognostic factors in patients with colon cancer. Consistent with our results, previous studies have also demonstrated that advanced age, higher pathological T and tumor recurrence are associated with poor prognosis in patients with colon cancer (31–33). Notably, this study revealed that the results of the prognostic model were consistent with actual survival prognostic information in different groups based on hierarchical analysis of age, higher pathological T and tumor recurrence. Meanwhile, the model status was also been considered as an independent prognostic factor in patients with colon cancer. These results further showed that prognostic model had a significant predictive ability for the prognosis of colon cancer.
In this study, the prognostic model was constructed based on the 12-prognostic gene signature (including 12 DEGs, ADORA3, CPA3, CPM, EDN3, FCRL2, MFNG, NAT1, PCSK5, PPARGC1A, PRRX2, TNFRSF17, and WDR78). Specifically, adenosine receptor A3 (ADORA3) protein encoded by ADORA3 gene is G-protein-coupled receptor that are implicated in inflammatory and immunological responses as well as cancer growth in various diseases through influencing nucleotide metabolic process (34–36). Increasing evidence has proved that ADORA3 is overexpressed in several cancers, including breast cancer (37), thyroid cancer (38), bladder cancer (39), and colon cancer (40) and functions as a tumor promoter (41). Carboxypeptidase A3 (CPA3) as a member of the CPA family of zinc metalloproteases is released by mast cells and may be involved in the inactivation of venom-associated peptides and the degradation of endogenous proteins (42). Previous study has shown elevated expression of CPA3 in asthma (43) and anaphylactic shock (44); however, few studies have investigated the role of CPA3 in cancers. CPM is also an arginine/lysine CP and exerts important roles in angiogenesis, proliferation, and apoptosis through modulating chemokines or kinins in cancer cells (45). Notably, recent study reports that CPM/Src-FAK pathway is involved in the cell migration and invasion in colon cancer (46). Endothelin 3 (END3) is reported to participate in the progression of several cancers, such as malignant melanoma (47), cervical cancer (48), and colon cancer (49). Fc Receptor Like 2 (FCRL2) is a member of the immunoglobulin receptor superfamily that is involved in the development of lymphoblastic leukemia by immunomodulators of B cell function (50–52). Inherited polymorphism in the acetyltransferase 1 gene (NAT1) increases the risk of colorectal adenocarcinoma (53). Manic fringe (MFNG) is reported to exhibit anti-tumor effects in lung cancer (54). Peroxisome proliferator-activated receptor-γ coactivator 1-α (PPARGC1A) can contribute to tumor growth and metastasis in several cancers (55, 56). In addition, studies have suggested that both paired related homeobox 2 (PRRX2) (57, 58) and tumor necrosis factor receptor superfamily member 17 (TNFRSF17) (59, 60) are associated with several cancers, while proprotein convertase subtilisin/kexin type 5 (PCSK5) and WD repeat domain 78 (WDR78) have not been reported to be involved in cancers. Thus, the functions of these genes in colon cancer should be further investigated.