Next we mapped the location of the most significant SNP of the 831 cis- associated genes from the single population analysis with respect to gene promoters, coding sequences, and conserved non-coding sequences. We observed significant excess of associated SNPs (relative to those tested) in promoters and coding sequences (Fisher's Exact test; P = 8.94 × 10−24, P = 4.52 × 10−12, respectively), and under-representation in conserved non-coding sequences (CNCs; P= 0.00358). The first two signals are quite expected and partly confounded since most of the causal variants are found in the genic regions, so SNPs in LD with causal variants may also map to coding sequences. Not surprisingly, we observed enrichment of associated SNPs in regions that align in multiple mammals and as far as fish or chicken (Fisher's exact test, P = 10−5. The apparent contradiction between enrichment in conserved nucleotides and deficit in CNCs is probably due to the fact that the majority of the signal for conserved nucleotide comes from exonic sequences that are close to the TSS where the associations are mainly found. The deficit of
