For network construction, we used bicor correlation and the following parameters: (1) mergeCutHeight set to 0.3 for the dentate gyrus and default values for the caudate nucleus, DLPFC and hippocampus; and (2) minModuleSizeset to 30 for the dentate gyrus and default values for the caudate nucleus, DLPFC and hippocampus. We set all other parameters to default values. The coexpression network was made using Pearson correlation values for the caudate nucleus (117 samples; 19,883 genes), dentate gyrus (46 samples; 18,747 genes), DLPFC (121 samples; 20,070 genes) and hippocampus (128 samples; 19,794 genes). We determined significant associations with ancestry using a linear model that correlated ancestry proportions (see ‘Estimation of genome-wide admixture levels’) with module eigengenes.
