Transcriptome sequencing uncovers a three–long noncoding RNA signature in predicting breast cancer survival

Long noncoding RNAs (lncRNAs) play a crucial role in tumorigenesis. The aim of this study is to identify lncRNA signature that can predict breast cancer patient survival. RNA expression data from 1064 patients were downloaded from The Cancer Genome Atlas project. Cox regression, Kaplan-Meier, and receiver operating characteristic (ROC) analyses were performed to construct a model for predicting the overall survival (OS) of patients and evaluate it. A model consisting of three lncRNA genes (CAT104, LINC01234, and STXBP5-AS1) was identified. The Kaplan-Meier analysis and ROC curves proved that the model could predict the prognostic survival with good sensitivity and specificity in both the validation set (AUC = 0.752, 95% confidence intervals (CI): 0.651-0.854) and the microarray dataset (AUC = 0.714, 95%CI: 0.615-0.814). Further study showed the three-lncRNA signature was not only pervasive in different breast cancer stages, subtypes and age groups, but also provides more accurate prognostic information than some widely known biomarkers. The results suggested that RNA-seq transcriptome profiling provides that the three-lncRNA signature is an independent prognostic biomarker, and have clinical significance. In addition, lncRNA, miRNA, and mRNA interaction network indicated lncRNAs may intervene in breast cancer pathogenesis by binding to miR-190b, acting as competing endogenous RNAs.