or cerebellar tissues compared with non-brain tissues. We also compared Cortex-EUR cis-eQTLs with eQTLGen14 (n = 31,684; blood-based, majority EUR ancestry), which supported the low agreement observed in GTEx blood. Of the overlapping eQTLs, 25% had an opposite allelic effect (AC = 75%, Rb = 0.52 and π1 = 0.83; Supplementary Fig. 15 and Supplementary Table 7) (ref. 16), which represents an increase over GTEx and suggests that many of the eQTLs are tissue-dependent. Combined, these results suggest that additional tissue- or ancestry-specific eQTLs can be identified when sample sizes increase. For instance, opposite effects may happen if two causal variants reside on the same haplotype but are specific for different tissues17, requiring large sample sizes for disentanglement. By revealing eQTLs with opposite allelic effects, our results highlight the relevance of tissue-dependent eQTL mapping to accurately assess the directionality of eQTLs17.
