DeGNServer:Deciphering Genome-Scale Networks through High Performance Reverse Engineering Analysis

Analysis of genome-scale Gene Networks (GNs) from large-scale gene expression profiles opens the door to uncover new biological knowledge. However, inferring genome-scale Gene Network (GN) from large-scale gene expression data and subsequent functional module mining are very computational intense tasks; therefore it requires both efficient algorithms and parallel computing engineering in order to enable and empower Genome-scale Gene Network analysis. Context likelihood of relatedness (CLR) method [1] based on the mutual information for scoring the similarity of gene pairs is one of the most accurate methods to infer gene networks. But it is computational unfeasibility to decipher a genome-wide network with large genomes, such as many plant genomes, with large-scale gene expression profiles, on a single computer due to limits on memory and CPU capacity. DeGNServer is a high performance web server that is capable of constructing genome-scale networks and further mining sub-networks from large genome-scale expression profiles for the species with large genomes/large number of genes:

The major feature includes:

• DeGNServer integrates six proven association methods(Spearman rank correlation,Pearson correlation,Mutual-information, Maximum information coefficient, Kendall rank correlation,Thei-Sen Estimator) for co-expression GN analysis and further utilize Context Likelihood of Relatedness approach for gene network analysis. In order to enable and empower genome-scale GN analysis, all algorithm have been implemented and deployed on our in-house parallel computing platform, namely BioGrid, which has over dedicated 700 CPU cores;

• Subnetwork identification and visualzation based on community structure mining methods;