by the author of the Sherlock software (Xin He, personal communication). We also slightly modified the Sherlock source code, omitting the exclusion criterion for SNPs (and genes whose expression is associated with those SNPs) that could not be found in the 1000 Genomes data, which encompassed only 49,612 [4.4%] of the 1,127,447 eSNPs also found in the PGC SCZ2 GWAS data 3. Default Sherlock parameters (priors) were used, except for setting the number of individuals in which the eQTL were discovered to N = 467, setting a 1% prevalence for SCZ, and setting the PGC SCZ2 GWAS primary meta-analysis cohort sizes (35,476 cases and 46,839 controls). For input allele frequencies, we used the frequencies estimated from the 46,839 GWAS controls. Also, instead of using the minor allele frequency, we used the “risk” allele frequency at each SNP; i.e., the allele at higher frequency in cases. This ensures that, for significantly associated SNPs, the minor or major allele is appropriately chosen for likelihood calculations based on the direction of risk. Still, for most SNPs in the genome, those not clearly associated with SCZ, the choice of major or minor allele is essentially random and unbiased.
