For internal validation of global ancestry-associated differential expression features (that is, gene, transcript, exon and junction), we performed differential expression analysis with a combination of BAs and WAs using mash. As with ‘Global ancestry-associated differential expression analysis’, we determined the effect size and s.e. of the effect size using limma-voom modeling. We replaced the continuous variable genetic ancestry with the binary, self-reported race. Additionally, we selected individuals with limited admixture by including: (1) Black Americans with African genetic ancestry ≥ 0.8; and (2) WAs with European genetic ancestry > 0.99. To limit the influence of the larger sample size compared to ‘Global ancestry-associated differential expression analysis’, we randomly sampled ten times without replacement to approximate the sample size of the admixed BA-only analysis. After extraction of the effect sizes and s.e., we implemented mash modeling for each feature across brain regions as described in the ‘Global ancestry-associated differential expression analysis’ section.
