Skip to main content


Fig. 5 | BMC Genomics

Fig. 5

From: Genome-wide enhancer annotations differ significantly in genomic distribution, evolution, and function

Fig. 5

Enhancers identified by different methods differ in functional attributes. The 9 kb region on human chromosome 1 containing genetic variants associated with LDL cholesterol levels and MI in GWAS and the causal SNP (rs12740374). Here, the region containing the casual SNP is predicted to be an enhancer by four of the seven methods. GWAS tag SNPs are colored in red and LD blocks are shown with a horizontal line. (b) The 60 kb region of human chromosome 9 containing loci associated with coronary artery disease (CAD) in GWAS. Two of the associated variants (rs10811656 and rs4977757) have been shown to contribute to CAD risk. However, the enhancer annotations in this region are generally non-overlapping and do not highlight either functional variant. (c) Few GWAS SNPs overlap an enhancer; the colored bars represent the number of methods that identified the region as an enhancer. The majority of these variants are not predicted as enhancers, and very few GWAS variants are overlap enhancers from multiple methods. The conclusions are similar when considering variants in high LD (r2 > 0.9) with the GWAS tag SNPs in liver (Liver LD; Additional file 1: Figure S8). The pattern is also similar when limiting to SNPs associated with liver or heart related phenotypes (Liver Specific, Heart Specific). When considering the SNP in each LD block with the maximum number of enhancer overlaps there is still a large percentage of SNPs supported by none or only one method (Liver Max). This demonstrates that the situation illustrated in panel B is very common. (d) Among all eQTL that overlap at least one enhancer, the majority is supported by only a single method. This holds for LD- expanded and context-specific sets (Liver LD, Liver Specific, Heart Specific; Additional file 1: Figure S8). Many variants remain unique to a single method, even when limiting to the variant in each LD block overlapping the maximum of enhancer sets (Liver Max). These trends are similar to what is seen for GWAS SNPs in (c). (e) Enhancer sets from the same biological context have different functional associations. We identified Gene Ontology (GO) functional annotations enriched among genes likely to be regulated by each enhancer set using GREAT. The upper triangle represents the pairwise semantic similarity for significant molecular function (MF) GO terms associated with predicted liver enhancers. The lower triangle shows the number of shared MF GO terms in the top 30 significant hits for liver enhancer sets. Results were similar when using enhancer-gene target predictions from JEME (Additional file 1: Figure S9–10)

Back to article page