The method for constructing scale-free networks by WGCNA has been described in previous studies [41, 93]. The gene co-expression networks were constructed by using the WGCNA v1.36 package in R environment (v3.02). In order to construct gene modules, pair-wise Pearson correlation coefficients were first calculated between all differentially expressed transcripts to generate a signed similarity matrix selecting for positive correlations only. To emphasize (weight) stronger correlations at the expense of weaker correlations, the signed similarity matrix was then raised to the lowest power, β = 14, that approximated a scale-free network topology (R2>0.80), to generate an adjacency matrix. Following this, a topological overlap measure (TOM) was calculated, which assessed transcript interconnectedness. A dissimilarity measure was calculated from the TOM and was subsequently used for average linkage hierarchical clustering. Module definition parameters included a minimum module size of 35 genes and a minimum module merge height of 0.8 (default parameter).
