Fig. 1

Identify and characterize sequence-conserved human enhancer-like elements. a Schematic of identifying ELEs (NC) and subsets with low (LC), moderate (MC) and high (HC) levels of sequence conservation. b Numbers of context-specific ELEs across 105 contexts. c, d Length (c) and evolutionary conservation score (d) distributions for NC, LC, MC, and HC ELEs. For b and d, each point denotes a median and each line denotes an interquartile range. e Numbers of overlapping pairs between ELEs and ENCODE cCREs. dELS and pELS, distal and proximal enhancer-like signature, respectively. PLS, promoter-like signature. DNH3, signature marked by DNase and H3K4me3. f Percentages of EUR and EAS common SNPs inside omnibus ELEs. g Percentages of EUR common SNPs inside context-specific ELEs. Each point denotes a context. h The heatmap shows the maximum correlations of SNP annotations based on NC context-specific ELEs from two context groups. BN, bone. CB, cancer (blood). CO, cancer (other). CV, cardiovascular. CN, connective. EN, endocrine. EP, epithelial. GI, gastrointestinal. IM, immune. MS, muscle. NR, neural. OM, omnibus. OT, other tissues (kidney, liver, lung). RP, reproductive. SK, skin. SC, stem cell. The scatter plot shows the pairwise correlations of SNP annotations based on NC and HC context-specific ELEs, where each point denotes a pair of contexts. i Correlations between a known SNP annotation and an ELE-based SNP annotation. Each point denotes a pair of an ELE-based SNP annotation and one of the 96 known SNP annotations. For h–i, the correlation of two binary annotations is quantified by Cramér’s V (from 0 to 1; \(V=0\): no correlation; \(V=1\): complete correlation). The correlation of a binary and a quantitative annotation is quantified by Pearson’s R (from –1 to 1; \(R=0\): no correlation; \(|R|=1\): complete correlation). For f–i, dashed lines have intercept 0 and slope 1