Differential expression between SCZ cases and the controls was assessed (Supplementary Fig. 2, bottom panel) utilizing the limma package in R, with the following inputs: the voom-normalized gene expression matrix, the voom precision weights matrix corresponding to the values in the expression matrix, and the final “covariate model”. Note that the expression matrix we utilized contained data for 592 samples in total (SCZ cases, controls, and AFF cases), and 16,423 genes passing the expression-level threshold of > 1 CPM in > 50% of the samples. The rationale for including ~50 AFF, even though they were not analyzed for differential expression, was to: a) increase statistical power during linear modeling of covariates (such as age, RIN, PMI, etc.); and b) place the expression data for the AFF cases on the same scale as for the SCZ and control samples for the sake of simplicity.
