Identifying candidate genes near sub-threshold loci using activity correlation across human tissues

From the Roadmap Epigenomics Project, we were able to obtain matching 'strong' enhancer annotations and RNA-seq data for 59 of the 127 tissues, including LV. For each LV enhancer, we considered all genes with expression ≥1 RPKM in LV and in vitro differentiated human cardiomyocytes and distance within +/-500 kb as potential targets. We then split the RNA-seq data for the 59 tissues into two groups, depending on whether the enhancer is present or absent in each tissue, and applied a one-sided Mann-Whitney U test to ask whether each potential target gene showed significantly greater expression in tissues where the enhancer was active. Genes differentially expressed between tissues with active and inactive enhancers (p<0.05) were considered computationally-determined potential target genes. For determining targets of sub-threshold enhancers, we first filtered our set of sub-threshold enhancers to remove those unlikely to be associated with QT interval. To do this, we excluded sub-threshold SNPs if the -log(p-value) was lower than 80% of the -log(p-value) of the most statistically significant SNP in LD (r²>0.2), as these are unlikely to be causal.