To provide a more global network analysis, a weighted gene co-expression network analysis (WGCNA) was also conducted (Zhang and Horvath, 2005), using the Bioconductor (Gentleman et al., 2004) package WGCNA (Langfelder and Horvath, 2008, 2012) within R (R: A language and environment for statistical computing Ver 2.15.0; R Foundation for Statistical Computing, 2013). Briefly, gene expression data of named genes were rank-ordered according to their ascending p-values obtained from traditional t-testing of the two experimental groups. For WGCNA, default values, including the use of the power function with power β, were used for all functions with the exception that signed correlation coefficients were used. A cutoff of p ≤ 0.10 was used to select genes for enrichment in networks that met the criteria of legitimacy for scale free topology (Zhang and Horvath, 2005). Resultant modules were tested for enrichment of various categories of genes using Fisher's Exact Test (p ≤ 0.05) plus enrichment by cell type (Cahoy et al., 2008). Modules that satisfied these two enrichment criteria were tested for significant (p ≤ 0.01) GO biologic process categories (Harris et al., 2004; Ashburner et al., 2000). Only significant GO categories with 40 or more genes are presented.
