(A and B) Mutual-information (MI) coexpression networks for HC (A; 962 edges) and SCZ (B; 234 edges) groups, showing 65 high-degree nodes selected from networks (HC: 16,041 nodes, 3.54 million edges; SCZ: 16,041 nodes, 1.73 million edges). Circles represent CGs (blue); diamonds represent TEs (orange for L1s, gold for HERVs). Node size is proportional to degree in the network. (C) Bar plots of genome-wide MI edge counts by transcript pair type for HCs (green) and SCZ (purple). (D) Box-and-whisker plots of node degree for genes, L1s, and HERVs in the HC and SCZ networks. Boxes span 25th–75th percentiles; center lines indicate medians; whiskers extend to 1.5× the interquartile range. (E) Volcano plot of differential expression (SCZ vs. HC) showing log₂ fold change (FC) versus –log₁₀ adjusted P value for genes (blue), L1 elements (orange), and HERVs (gold). Dashed lines indicate thresholds of adjusted P < 0.01 and |log₂FC| > 0.5. (F) Heatmap of the top 50 upregulated and downregulated transcripts ranked by combined score (–log₁₀[adjusted P] × |log₂FC|). Color scale represents row-wise z scores of variance-stabilized transformed counts. Differential expression was assessed using DESeq2 with correction. Node degree differences (D) were tested by nonparametric bootstrap (100,000 iterations).