By assaying gene expression using the Affymetrix exon array platform, we were able to differentiate between cis-eQTL signals regulating all expressed exons in a gene from those regulating a subset of exons. Consistent with growing evidence for the widespread nature and functional importance of alternative splicing in the human brain, we found that most of our cis-eQTL signals were only apparent using exon-level data. Only 29.6–39.2% of our cis-eQTL signals were reflected in a significant gene-level signal, but almost all were represented by at least one exon-level signal (Supplementary Table 5). The majority of signals would therefore have not been identified without exon-level information. This finding is not explained by differing false positive rates among groups of cis-eQTL signals, as replication rates for gene-level and exon-level signals were similar (Supplementary Table 3).
