Identification and validation of a novel signature as a diagnostic and prognostic biomarker in colorectal cancer.

BACKGROUND

Colorectal cancer (CRC) is one of the most common malignant neoplasms worldwide. Although marker genes associated with CRC have been identified previously, only a few have fulfilled the therapeutic demand. Therefore, based on differentially expressed genes (DEGs), this study aimed to establish a promising and valuable signature model to diagnose CRC and predict patient's prognosis.

METHODS

The key genes were screened from DEGs to establish a multiscale embedded gene co-expression network, protein-protein interaction network, and survival analysis. A support vector machine (SVM) diagnostic model was constructed by a supervised classification algorithm. Univariate Cox analysis was performed to construct two prognostic signatures for overall survival and disease-free survival by Kaplan-Meier analysis, respectively. Independent clinical prognostic indicators were identified, followed by univariable and multivariable Cox analysis. GSEA was used to evaluate the gene enrichment analysis and CIBERSORT was used to estimate the immune cell infiltration. Finally, key genes were validated by qPCR and IHC.

RESULTS

In this study, four key genes (DKC1, FLNA, CSE1L and NSUN5) were screened. The SVM diagnostic model, consisting of 4-gene signature, showed a good performance for the diagnostic (AUC = 0.9956). Meanwhile, the four-gene signature was also used to construct a risk score prognostic model for disease-free survival (DFS) and overall survival (OS), and the results indicated that the prognostic model performed best in predicting the DFS and OS of CRC patients. The risk score was validated as an independent prognostic factor to exhibit the accurate survival prediction for OS according to the independent prognostic value. Furthermore, immune cell infiltration analysis demonstrated that the high-risk group had a higher proportion of macrophages M0, and T cells CD4 memory resting was significantly higher in the low-risk group than in the high-risk group. In addition, functional analysis indicated that WNT and other four cancer-related signaling pathways were the most significantly enriched pathways in the high-risk group. Finally, qRT-PCR and IHC results demonstrated that the high expression of DKC1, CSE1L and NSUN5, and the low expression of FLNA were risk factors of CRC patients with a poor prognosis.

CONCLUSION

In this study, diagnosis and prognosis models were constructed based on the screened genes of DKC1, FLNA, CSE1L and NSUN5. The four-gene signature exhibited an excellent ability in CRC diagnosis and prognostic prediction. Our study supported and highlighted that the four-gene signature is conducive to better prognostic risk stratification and potential therapeutic targets for CRC patients.