Five-mRNA Signature for the Prognosis of Breast Cancer Based on the ceRNA Network

BACKGROUND: The purpose of this study was to investigate the regulatory mechanisms of ceRNAs in breast cancer (BC) and construct a new five-mRNA prognostic signature. METHODS: The ceRNA network was constructed by different RNAs screened by the edgeR package. The BC prognostic signature was built based on the Cox regression analysis. The log-rank method was used to analyse the survival rate of BC patients with different risk scores. The expression of the 5 genes was verified by the GSE81540 dataset and CPTAC database. RESULTS: A total of 41 BC-adjacent tissues and 473 BC tissues were included in this study. A total of 2,966 differentially expressed lncRNAs, 5,370 differentially expressed mRNAs, and 359 differentially expressed miRNAs were screened. The ceRNA network was constructed using 13 lncRNAs, 267 mRNAs, and 35 miRNAs. Kaplan-Meier (K-M) methods showed that two lncRNAs (AC037487.1 and MIR22HG) are related to prognosis. Five mRNAs (VPS28, COL17A1, HSF1, PUF60, and SMOC1) in the ceRNA network were used to establish a prognostic signature. Survival analysis showed that the prognosis of patients in the low-risk group was significantly better than that in the high-risk group (p = 0.0022). ROC analysis showed that this signature has a good diagnostic ability (AUC = 0.77). Compared with clinical features, this signature was also an independent prognostic factor (HR: 1.206, 95% CI 1.108−1.311; p < 0.001). External verification results showed that the expression of the 5 mRNAs differed between the normal and tumour groups at the chip and protein levels (p < 0.001). CONCLUSIONS: These ceRNAs may play a key role in the development of BC, and the new 5-mRNA prognostic signature can improve the prediction of survival for BC patients.