for nonsynonymous and loss-of-function mutations, were combined by taking the minimum P value after Bonferroni correction for the 2 tests. d) Rare variants: PLINK/Seq and SMP5 (http://atgu.mgh.harvard.edu/plinkseq) were used to assess whether the exome-sequenced SCZ cases exhibited a burden of rare singleton variants (observed just once in the entire cohort of ~5,000 individuals) in the differentially expressed genes, as compared to controls. Enrichment statistics for the differentially expressed set (the sum of gene burden statistics) were calculated via permutation that controlled for any exome-wide case-control differences, residual linkage disequilibrium among rare variants in nearby genes, and differences between cases and controls arising from ancestry (based on exome-wide identity-by-state [IBS]), experimental batch, and gender. Case burden in the differentially expressed genes was tested for either nonsynonymous variants (comprised of loss-of-function variants and missense variants predicted in silico as deleterious by each of five different algorithms5), or just the smaller set of loss-of-function variants. Looking only at differentially expressed DLPFC genes, there were 236 genes with one or more singleton loss-of-function variants and a total of 440 genes harboring singleton damaging nonsynonymous variants. 10,000 permutations were used for each test. Again, the two tests were combined by choosing the minimal P value
