To identify trans-eQTLs, we performed a limited analysis to reduce the multiple-testing burden by focusing on 228,819 variants with a known interpretation. This set constituted variants that were either previously associated with traits, having a GWAS P < 5 × 10−8 in the IEU OpenGWAS database89 or EBI GWAS catalog90 on 3 May 2020, and additional neurological traits (Supplementary Table 17) or that showed an association with q-value < 0.05 in any of our discovery cis-eQTL analyses (including non-primary associations identified in the iterative conditional analysis). To maximize power, we combined the Cortex-EUR and Cortex-AFR datasets but excluded the ENA cohort due to the potential for genotypes of poorer quality (n = 2,759; Supplementary Note). For this dataset, we also repeated the cis-eQTL analysis (Supplementary Table 2) and normalization approach, including the selection of optimal number of 100 PCs to regress out (Supplementary Fig. 7). We assessed those combinations of SNPs and genes where the SNP–TSS distance was >5 Mb or where the gene and SNP were on different chromosomes as trans-eQTLs. To determine significance, we employed a previously used
