To confirm cell type-specific eQTL effects identified in Cortex-EUR, we used three replication datasets: Cortex-AFR and the snRNA-seq datasets Bryois et al.24 and ROSMAP4. For the Cortex-AFR replication, we applied the same cell-type prediction and Decon-QTL interaction analysis as for Cortex-EUR. Over the ieQTLs significant in Cortex-EUR, we calculated BH-FDR estimates and deemed ieQTLs with a BH-FDR < 0.05 as significant. Given that Decon-QTL does not return any standard errors, we predicted β and standard errors using the sample size, MAF, interaction β and interaction P value97 to calculate the Rb metrics. For the ROSMAP dataset, encompassing 80,660 single-nucleus transcriptomes from the prefrontal cortex of 48 individuals with varying degrees of AD pathology23, we re-processed the expression matrix to create a pseudo-bulk expression matrix for each broad cell type and subsequently mapped cis-eQTLs using the same procedure as the trans-eQTL analysis in bulk data (Supplementary Note and Supplementary Figs. 39,40). To correct for multiple testing, we confined the analysis to only test for primary cis- or trans-eQTLs that had a significant interaction with one or more cell types in MetaBrain
