We started by curating two classes of gene sets for analyzing the differential expression data: 1) a small group of pathways and gene sets previously implicated in genome-wide genetic studies of schizophrenia (“hypothesis-driven”), and 2) a collection of thousands of “hypothesis-free” gene sets from large databases that would allow us to potentially characterize novel biology arising in brain expression related to schizophrenia. We considered each of these classes independently for multiple test correction owing to their dissimilar goals. Hypothesis-driven: This collection consisted of 12 sets of genes previously implicated in the literature of schizophrenia genetics, including: a) all genes within 20 kb of 108 GWAS loci3, b) genes sitting under rare SCZ-associated CNV88, and c) nonsynonymous and loss-of-function de novo mutations discovered from exome-sequencing of schizophrenia probands and their parents7,89–92; note that these correspond to the data described above, where all genes in associated regions are simply lumped together as a single gene set (losing the important distinction that some loci bear many more genes than others). In addition, we added gene sets previously shown to be enriched for genetic
