Progress in ovarian cancer treatment lags behind other tumor types. With diagnosis usually at an advanced stage, there is a high demand for reliable prognostic biomarkers capable of the selection of effective chemo- and targeted therapies. Our goal was to establish a large-scale transcriptomic database and use it to uncover and rank survival-associated genes. Ovarian cancer cohorts with transcriptome-level gene expression data and clinical follow-up were identified from public repositories. All samples were normalized and entered into an integrated database. Cox univariate survival analysis was performed for all genes and was followed by multivariate analysis for selected genes involving clinical and pathological variables. False discovery rate was computed for multiple hypothesis testing and a 1% cutoff was used to determine statistical significance. The complete integrated database comprises 1816 samples from 17 datasets. Altogether, 2468 genes were correlated to progression-free survival (PFS), and 704 genes were correlated with overall survival (OS). The most significant genes were WBP1L, ASAP3, CNNM2, and NCAPH2 for progression-free survival and CSE1L, NUAK1, ALPK2, and SHKBP1 for overall survival. Genes significant for PFS were also preferentially significant for predicting OS as well. All data including HR and p values as well as the used cutoff values for all genes for both PFS and OS are provided to enable the ranking of future biomarker candidates across all genes. Our results help to prioritize genes and to neglect those which are most likely to fail in studies aiming to establish new clinically useful biomarkers and therapeutic targets in serous ovarian cancer.
Keywords: Cox regression; Gene arrays; Gene expression; Survival.
© 2023. The Author(s).