MiRank: A bioinformatics tool for gene/miRNA ranking and pathway profiling with TCGA-KEGG data sets
The Cancer Genome Atlas (TCGA) provides researchers with clinicopathological data and genomic characterizations of various carcinomas. These data sets include expression microarrays for genes and microRNAs -- short, non-coding strands of RNA that downregulate gene expression through RNA interference -- as well as days_to_death and days_to_last_followup fields for each tumor sample. Our aim is to develop a software tool that screens TCGA data sets for genes/miRNAs with functional involvement in specific cancers. Furthermore, our computational pipeline is intended to produce a set of visualizations, or profiles, that place our screened outputs in a pathway-centric context. We accomplish our 'screening' by ranking genes/miRNAs by the correlation of their expression misregulation with differential patient survival. In other words, if a gene/miRNA is consistently misregulated in patients with poor survival rates and, on the other hand, is expressed more 'normally' in patients with longer survival rates, then it is ranked highly; if its misregulation has no such correlation with good/bad survival in patients, then its rank is low. Our pathway profiling pipeline produces several outputs, which allow us to examine the functional roles played by highly ranked genes discovered by our screening. Running the OV (ovarian serous cystadenocarcinoma) data set through our analysis pipeline, we find that several highly ranked pathways and functional groups of genes (VEGF, Jun, Fos, etc.) have already been shown to play some part in the development of epithelial ovarian carcinomas. We also observe that the dysfunction of the Wnt signaling pathway, which regulates cell-fate specification and progenitor cell differentiation, has a disproportionate impact on the survival of ovarian cancer patients.