A prognostic risk model for ovarian cancer based on gene expression profiles from gene expression omnibus database

This study explored prognostic genes of ovarian cancer and built a prognostic model based on these genes to predict patient's survival, which is of great significance for improving treatment of ovarian cancer. GSE26712 dataset was downloaded from Gene Expression Omnibus database as training set, while OV-AU dataset was downloaded from ICGC website as validation set. All genes in GSE26712 were analyzed by univariate Cox regression, Lasso regression, and multivariate Cox regression analyses. Then prognosis-related feature genes were screened to construct a multivariate risk model. Meanwhile, Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis was performed on samples in the high/low-risk groups using Gene Set Enrichment Analysis (GSEA) software. Finally, survival curve and receiver operating characteristic curve were drawn to verify the validity of the model. Ten feature genes related to prognosis of ovarian cancer were obtained: CMTM6, COLGALT1, F2R, GPR39, IGFBP3, RNF121, MTMR9, ORAI2, SNAI2, ZBTB16. GSEA enrichment analysis showed that there were notable differences in biological pathways such as gap junctions and homologous recombination between the high/low-risk groups. Through further verification of training set and validation set, the 10-gene prognostic model was found to be effective for the prognosis of ovarian cancer patients. In this study, we constructed a 10-gene prognostic model which predicted the prognosis of ovarian cancer patients well by integrating clinical prognostic parameters. It may have certain reference value for subsequent clinical treatment research of ovarian cancer patients and help in clinical treatment decision-making.