We next evaluated whether SCZ cases versus controls differed in their expression levels per gene. Following normalization of read counts for each gene, a weighted linear regression adjusting for known covariates was performed (Supplementary Figs. 2 and 4). Analysis of the distribution of P values for the 16,423 genes was tested for a mixture of disease-associated and null distributions for 25 cases and 25 controls and suggests that approximately 44% of genes are perturbed in SCZ; this excess of low P values disappears when case and control labels are permuted. While polygenic inheritance, where many genes are affected but to a small degree3, could explain this result, treatment and environmental factors also likely play a role. Without imposing a threshold on the magnitude of fold change in mean expression between SCZ and controls, we find 693 genes to be differentially expressed after correction for multiple testing (FDR ≤ 5%), 332 up-regulated and 361 down-regulated (Fig. 5A, Supplementary data file 3). All had modest fold changes (Fig. 5B), with a mean of 1.09 and range 1.03–1.33 (inverting down-regulated expression ratios). As
