OncoRank: A pan-cancer method of combining survival correlations and its application to mRNAs, miRNAs, and lncRNAs

Abstract

OncoRank adopts a method for finding recurrent miRNA-target interactions to find genes with consistent relationships to patient survival across cancers. Genes are first ranked in each cancer by their Cox coefficients, and these ranks are then combined by applying Fisher's method. Using ranks instead of the raw coefficients or p-values allows each cancer to be weighted equally and prevents bias from cancers with large numbers of patients. OncoLnc (http://www.oncolnc.org) is a newly available resource for Cox coefficients and utilizes data from 21 cancers in The Cancer Genome Atlas. Using this resource I applied OncoRank to mRNAs, miRNAs, and lncRNAs and in each case found consistently harmful or protective genes. These genes may be members of central cancer pathways and should be of interest to cancer researchers.