OBJECTIVE : Approximately, 30% patients after radical prostatectomy (RP) will undergo post-operative biochemical recurrence (BCR). Present stratification method by TNM staging and Gleason score was not adequate to screen high-risk patients. In this study, we intended to identify a novel set of differentially expressed gene (DEG) signature that can predict BCR after RP. UNASSIGNED : 358 patients after RP with follow-up data were extracted from The Cancer Genome Atlas (TCGA), among which 61 patients had undergone BCR. Key DEGs were confirmed by the intersection of GSE35988 and TCGA_PCa dataset, and their gene expression data were also extracted from TCGA_PCa dataset. Kaplan-Meier plot and Cox proportion hazard regression model were applied to assess the relationship between risk score and survival outcome (BCR). RESULTS : 310 DEGs were confirmed in two prostate cancer dataset. 6 DEGs (SMIM22, NINL, NRG2, TOP2A, REPS2, and TPCN2) were selected to construct a risk score formula. The risk score was a powerful predictive factor independent of TNM stage (HR 3.045, 95% CI 1.655-5.602, p < 0.001). CONCLUSIONS : In this study, a novel 6-gene signature with robust predictive ability on post-operative BCR was constructed and 4 genes (SMIM22, NRG2, NINL and TPCN2) in the 6-gene signature were not reported to be associated with prostate cancer.