Identification of a three-gene signature in the triple-negative breast cancer

This work aimed to improve current prognostic signatures based on clinical stages in identifying high-risk patients of triple-negative breast cancer (TNBC), to allow patients with a high-risk score for specific treatment decisions. In this study, 396 TNBC samples from TCGA and GEO databases were included in genome-wide transcriptome analysis. The relationship between normalized gene expression values and survival data of patients was determined by Cox proportional hazards models in each dataset. The overlapped genes among all datasets were considered as a potential prognostic signature. The risk score was constructed based on individual genes and validated with three separate data sets and the combined dataset. Moreover, the Kaplan-Meier analysis including the log-rank test was performed to determine significantly statistical differences in overall survival. The association analysis between DNA methylation levels and gene expression levels of three genes was measured in the TCGA data set. In Cox proportional hazards model analysis, the result showed that potential protective genes included 564 genes in GSE25066 dataset, 1132 genes in GSE103091 dataset and 564 genes in TCGA dataset, potentially risky genes contained 1132 genes in GSE25066 dataset, 475 genes in GSE25066 dataset and 1115 genes in TCGA dataset. In all datasets, patients in high-risk groups showed worse prognosis than low-risk groups. Multivariate Cox regression analysis displayed that the 3-gene signature (DCAF4, UQCRFS1 and SS18) was an independent prognostic factor in TNBC. The AUC values of the 3-gene signature were 0.71, 0.73 and 0.77 in GSE25066, GSE103091 and TCGA dataset, respectively. The model, which combined the 3-gene signature (DCAF4, UQCRFS1 and SS18) with the tumor stage (pathological stage or pathological T stage), showed a stronger prognostic power for survival prediction. The 3-genes prognostic signature may be a useful biomarker for survival prediction in TNBC patients and may contribute to patient classification in the same tumor stages and individualized clinical treatment.