Identification of Biologically Significant Elements Using Correlation Networks in High Performance Computing Environments.

Network modeling of high throughput biological data has emerged as a popular tool for analysis in the past decade. Among the many types of networks available, the correlation network model is typically used to represent gene expression data generated via microarray or RNAseq, and many of the structures found within the correlation network have been found to correspond to biological function. The recently described gateway node is a gene that is found structurally to be co-regulated with distinct groups of genes at different conditions or treatments; the resulting structure is typically two clusters connected by one or a few nodes within a multi-state network. As network size and dimensionality grows, however, the methods proposed to identify these gateway nodes require parallelization to remain efficient and computationally feasible. In this research we present our method for identifying gateway nodes in three datasets using a high performance computing environment: quiescence in Saccharomyces cerevisiae, brain aging in Mus Musculus, and the effects of creatine on aging in Mus musculus. We find that our parallel method improves runtime and performs equally as well as sequential approach.