Ovarian cancer (OC) is the most malignant tumor in the female reproductive tract. Although abundant molecular biomarkers have been identified, a robust and accurate gene expression signature is still essential to assist oncologists in evaluating the prognosis of ovarian cancer patients. In the current study, 367 patients were adopted through The Cancer Genome Atlas (TCGA) database, and mRNA expression profiling was performed. Then, we used a gene set enrichment analysis (GSEA) to screen genes correlated with the epithelial-to-mesenchymal transition (EMT) process and identify these genes with the Cox proportional regression model. Six genes (TGFBI, SFRP1, COL16A1, THY1, PPIB, BGN) associated with overall survival (OS) were used to construct a risk assessment model, by which the patients were divided into high-risk and low-risk groups. The six-gene signature was identified as an independent prognostic biomarker of the OS of ovarian cancer patients via multivariate Cox regression analysis. Besides, the six-gene model was also validated significantly by Gene Expression Omnibus (GEO) database. In summary, we established a six-gene signature relevant to the prognosis of ovarian cancer, which might become a therapeutic target with clinical usefulness in the future.