eQTL lookups were carried out in the 13 brain tissues available in GTEx V7 [39], Brain xQTL (dorsolateral prefrontal cortex) [40] and BRAINEAC [41] databases, all of which had undergone QC by the individual studies. We did not perform additional QC on these data. In brief, GTEx used Storey’s q-value method to correct the FDR for testing multiple transcripts based upon the empirical P-values for the most significant SNV for each transcript [43, 42]. BRAINEAC calculated the number of tests per transcript and used Benjamini–Hochberg procedure to calculate FDR per transcript using a FDR < 1% as significant. BRAINxQTL used P < 8 × 10−8 as a cut-off for significance for any given transcript. SNVs that met the study specific significance and FDR thresholds, which were in LD (r2 > 0.8 in 1000 Genomes Europeans) with the top eQTL or the sentinel eQTL for a given tissue/transcript combination were considered significant. The genes implicated by these eQTL databases and/or coding changes (e.g., missense and nonsense SNVs) were put into ConsensusPathDB [44] to identify whether these genes were over-represented in any
