The coexpression network was constructed by means of the Weighted Gene Coexpression Network Analysis (WGCNA) (Iancu et al. 2012; Langfelder & Horvath 2008) using a consensus module approach followed by assessment of selection effects on network structure (Ando et al. 2015; Gill et al. 2010; Iancu et al. 2013a; Ideker & Krogan 2012). Cosplicing networks were constructed using a procedure we have termed CoSplicEx (Iancu et al. 2015). The procedure for CoSplicEx network construction was identical to that for the coexpression networks, except that the Pearson correlations were replaced with Mantel correlations (Iancu et al. 2015). In the coexpression network, the difference in correlation strength was evaluated utilizing the ‘var.test’ R function. To mitigate the computational load, we restricted the search to Pearson correlations between individual genes that differed by ≥0.5 before power transformation. For CoSplicEx edges, we implemented a permutation procedure that compared differences in correlation strengths between the selected groups with differences in correlation strengths between two randomly assigned groups. This general procedure has been used to quantify network rewiring in both genomic (Gill et al. 2010) and
