We investigated ancestry, brain region, dataset and tissue-type differences in cis-eQTLs. Agreement between ancestries was high: allelic concordance (AC) and correlation of effect-size (Rb) estimates were high when different ancestries were compared (Rb > 0.78, AC > 92.95%; Fig. 3c, Supplementary Fig. 12 and Supplementary Table 5). The proportion of estimated true-positives (π1) and correlation of allelic fold change (caFC) estimates between ancestries were lower, potentially due to differences in sample size (for example, Cortex-EUR versus Cortex-EAS, caFC = 0.55 and π1 = 0.29; conversely, caFC = 0.85 and π1 = 0.95; Supplementary Fig. 12). Similarly, different brain regions showed high overall agreement (Rb > 0.76, caFC > 0.65, and AC > 91%), with π1 estimates dependent on the sample size (0.39–0.95). Cerebellum was an exception and showed lower agreement with the cerebral brain regions (Fig. 3d,e and Supplementary Fig. 12). Despite the limited sample size, we identified 477 cis-eQTL genes that are significant in Cerebellum-EUR but not in Cortex-EUR (Supplementary Fig. 13), perhaps due to low expression in the cortex or because they are regulated by transcription factors that
