Figure 2

eQTL replication in brain data. Normalization by EDASeq (red bars) demonstrates that sample outlier removal improves eQTL replication and that whole gene annotation provides improved gene expression estimation than coding sequence (CDS) only. CQN normalization (green bars) provides slightly improved eQTL replication over EDASeq normalization and demonstrates the necessity of covariate inclusion in eQTL replication, particularly highlighting the necessity of accounting for unknown covariates. Per-gene outlier removal (PGOR, blue bars) does not hamper our ability to detect cis-eQTLs.