WGCNA and network terminology

WGCNA is a well-established technique for gleaning biologically relevant clusters of coexpressed and functionally related genes from microarray and sequencing data. WGCNA methods and terminology are summarized and defined in numerous manuscripts (Hilliard et al., 2012a; Zhang and Horvath, 2005; Dong and Horvath, 2007; Zhao et al., 2010; Yip and Horvath, 2007; Horvath, 2011). For the sake of convenience, we provide working definitions of network terms that we use throughout the manuscript. Definitions of greater detail are available in the manuscripts cited above.

Adjacency (a): The first step of network construction is to generate an adjacency matrix where A_ij = S_ij^β, where i and j are genes, S is the expression correlation across samples, and β is an empirically derived power to which the correlation is raised such that the resulting network approximates a scale free topology.

Connectivity (k): Connectivity is a measure of connectedness of a given gene, either in the context of its module (kIN) or the entire network (kTotal). Connectivity is defined as follows: ki=∑j=1Naij where i and j are genes, N is all of the genes in the module or network, and a is the adjacency between genes i and j.

Topological overlap: Adjacency is transformed to topological overlap as a method of calculating the interconnectedness (or similarity) between two nodes. Topological overlap is defined as follows: ωij=lij+aijmin{ki,kj}+1−aij and lij=∑u≠i,jaiuauj, where u represents all genes besides i and j. A and k are defined above.

Gene significance: The Pearson correlation between a gene’s expression profile and, in our work, a given behavioral metric.

Module eigengene: The first principal component of a module’s gene expression profile, a method of summarizing an entire module in one vector.

Module membership: The correlation between an individual gene expression profile and a module eigengene. Genes with high module membership tend to have high intramodular connectivity and are referred to as intramodular hubs. Of note, genes can have high module membership in more than one module.

Zsummary: Along with median rank, a term for quantifying preservation of gene coexpression patterns between two independent datasets (Langfelder et al., 2011), such as between juvenile and adult Area X or juvenile Area X and juvenile VSP. Zsummary is a composite preservation score defined as the average of Zdensity and Zconnectivity, which assess the preservation of connection strength among network nodes (e.g. Are strongly connected nodes in one network also strongly connected in the other?) and the connectivity patterns between nodes (e.g. Do the patterns of connection between specific nodes exist in both networks?), respectively, following permutation tests under the null hypothesis. Higher Zsummary scores indicate better preservation.