Gene expression based prediction of prognostic outcome in ovarian cancer

Abstract

Gene expression provides rich information. Successful application has made to predict prognosis of several cancers such as breast and colon. However, although ovarian cancer is the fifth leading death cancer to women, precise prediction of survival outcome is not available yet. Thus there is a still urgent need for optimized treatment decision. Recent studies made use of public gene expression data sources to predict the clinical outcome of ovarian cancer. Typically, two steps approach has tried. First step is figuring out significant genes by univariate Cox regression model. Second step is providing a statistic that will combine the effect of selected genes in terms of survival risk. One of drawback of the two steps approach is low reproducibility. Statistics for risk group classification built in the train set often fails to be validated when the statistic is applied to the data set. Applying the scheme to the RNAseq data from The Cancer Genome Atlas(TCGA) has shown that the classification results of the patient's prognosis was classified higher and lower risk patient of the patient's prognosis. We applied median standard to the classification of existing scheme and suggested other schemes for the successive work.