Abstract
Biology is entering an era of big data in which polygenic traits and diseases can be measured and mined with their molecular intricacies intact. For example, The Cancer Genome Atlas (TCGA) contains petabytes of molecular and clinical information representing a diverse array of tumors. This data is available for multi-omics analysis capable of improving our understanding of tumor etiology and identifying novel treatment options. Here we input TCGA data into a novel bioinformatics workflow, Knowledge-Independent Network Construction (KINC), to enhance tumor gene expression analysis at the transcriptome (steady-state RNA) level. KINC builds a network of condition-specific gene interactions where RNA transcripts are nodes and significant RNA co-expression dependencies are edges. Our results suggest that, compared to current techniques, KINC more selectively describes correlations between genes. We applied KINC to over 2000 TCGA tumor gene expression datasets spanning five human cancers – bladder cancer, ovarian cancer, thyroid cancer, lower grade glioma, and glioblastoma. With KINC, we found disease-specific gene modules that suggest known and novel commonalities between different cancers. These data were processed using the Open Science Grid (OSG) computing platform. OSG and other relevant cyberinfrastructure will be discussed.