There is no gene signature for predicting relapse and survival of cervical cancer with early stage currently. In this study, we investigate whether gene expression profiling of cervical cancer could be used to predict the prognosis of patient. A series of 100 primary cervical cancer patients who underwent radical hysterectomy between January 2001 and October 2006 were analyzed for gene expression profiles by using a custom oligonucleotide microarray containing probes for 1440 human tumor-related gene transcripts. Supervised analysis of gene expression data identified 19 genes that exhibited differential expression between cervical cancer and normal cervix. Then, all 100 patients were divided into the training (n=50) and testing sets (n=50). Using Cox regression and risk-score analysis, we identified a 7-gene (UBL3, FGF3, BMI1, PDGFRA, PTPRF, RFC4, and NOL7) signature for predicting relapse of patient in the training set. The 7-gene signature was validated by the testing set (sensitivity, 84.6%; specificity, 91.9%; positive predictive value, 78.6%; negative predictive value, 94.4%). Patients with high-risk 7-gene signature had poor relapse-free survivals (RFS) than patients with low-risk 7-gene signature in both training set (P=0.026) and testing set (P=0.042). Multivariate analysis showed that the FIGO stage and 7-gene signature are independent prognostic factors associated with RFS of cervical cancer patients. The 7-gene signature can predict cancer recurrence and survival of cervical cancer patients. This may have prognostic or therapeutic implications for the future management of cervical cancer patients.