Network statistics of genetically-driven gene co-expression modules in mouse crosses

• Main Authors: Marie-Pier eScott-Boyer, Benjamin eHaibe-Kains, Christian F Deschepper
• Format: Article
• Language: English
• Published: Frontiers Media S.A. 2013-12-01
• Series: Frontiers in Genetics
• Subjects: Genetics; Network Inference; mouse recombinant inbred strains; gene co-expression modules; chromosome domain
• Online Access: http://journal.frontiersin.org/Journal/10.3389/fgene.2013.00291/full

In biology, networks are used in different contexts as ways to represent relationships between entities, such as for instance interactions between genes, proteins or metabolites. Despite progress in the analysis of such networks and their potential to better understand the collective impact of genes on complex traits, one remaining challenge is to establish the biologic validity of gene co-expression networks and to determine what governs their organization. We used WGCNA to construct and analyze seven gene expression datasets from several tissues of mouse recombinant inbred strains (RIS). For six out of the 7 networks, we found that linkage to module QTLs (mQTLs) could be established for 29.3% of gene co-expression modules detected in the several mouse RIS. For about 74.6% of such genetically-linked modules, the mQTL was on the same chromosome as the one contributing most genes to the module, with genes originating from that chromosome showing higher connectivity than other genes in the modules. Such modules (that we considered as genetically-driven) had network statistic properties (density, centralization and heterogeneity) that set them apart from other modules in the network. Altogether, a sizeable portion of gene co-expression modules detected in mouse RIS panels had genetic determinants as their main organizing principle. In addition to providing a biologic interpretation validation for these modules, these genetic determinants imparted on them particular properties that set them apart from other modules in the network, to the point that they can be predicted to a large extent on the basis of their network statistics.