Long non-coding RNA-based signatures to improve prognostic prediction of breast cancer

Breast cancer (BC) is a disease of high mortality rate because of high malignant, while early diagnosis and personal management may make a better prognosis possible. This study aimed to establish and validate lncRNAs signatures to improve the prognostic prediction for BC. RNA sequencing data along with the corresponding clinical information of patients with BC were gained from The Cancer Genome Atlas (TCGA). Prognostic differentially expressed lncRNAs were obtained using differentially expressed lncRNAs analysis (P value <.01 and |fold change| > 2) and univariate cox regression (P value <.05). By applying least absolute shrinkage and selection operation (LASSO) Cox regression analysis along with 10-fold cross-validation, 2 lncRNA-based signatures were constructed in the training, test and whole set. A 14-lncRNAs signature and a 10-lncRNAs signature were built for overall survival (OS) and relapse-free survival (RFS) respectively in the 3 sets. BC patients were divided into high-risk groups and low-risk groups depended on median risk score value. Significant differences were found for OS and RFS between 2 groups in the 3 sets. The time-dependent receiver operating characteristic (ROC) curves analysis demonstrated that our lncRNAs signatures had better predictive capacities of survival and recurrence for BC patients as well as enhancing the predictive ability of the tumor node metastasis (TNM) stage system. These results indicate that the 2 lncRNAs signatures with the potential to be biomarkers to predict the prognosis of BC for OS and RFS.