A lymph node metastasis-related protein-coding genes combining with long noncoding RNA signature for breast cancer survival prediction.

Integrating protein-coding gene (PCG) with long noncoding RNA (lncRNA) expression profiles, our aim is to identify a multidimensional transcriptome model that can predict individual prognosis of patients with breast cancer (BRCA). After diverse bioinformatics and statistical analyses, we obtained gene expression profiles of 1,016 BRCA samples from Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) database, and constructed a prognostic signature, which is composed of four PCGs (TFCP2, LRRC75B, PROSER2, and STOML1) and one lncRNA (AL355592.1). In the training set, the multidimensional transcriptome signature could part patients with BRCA into two groups with different survival, defined as high- and low-risk group by Kaplan-Meier (KM) analysis (p < 0.001). In the other five validation datasets, the PCG-lncRNA signature showed a similar predictive performance in BRCA by KM (p < 0.05). The prognostic independence for the PCG-lncRNA model was verified by the multivariable Cox regression analysis. Because Chi-squared test showed the signature was associated with lymph node metastasis status, stratification analysis found that it could further subdivide lymph node metastasis status more precisely in BRCA. The function analysis suggested that the genes from the signature were associated with immunity-related pathways. In summary, we constructed a PCG-lncRNA signature, which could accurately predict survival in patients with BRCA.