Fig. 2

DMVs are hotspots of transcription factor (TF) binding sites. a The average distribution of TF binding sites (per kb) around human DMVs or CpG island (CGI) clusters based on ChIP-seq data from ENCODE [35]. Random sequences with the same lengths as DMVs in the genome were used as a control. CGI clusters were defined as described previously [14]. Super-enhancers were defined in human embryonic stem cells as previously described [37]. b UCSC Genome Browser snapshots of DNA methylomes for sperm [85] and various cell types [14] and TF binding density [35] in human DMVs near Foxa1 and Gata2. c The average distribution of TF bindings around transcription start sites (TSSs) in human [35] for promoters in DMVs or other promoters. d TF motif analysis of different regions in the human genome CGI in DMVs, non-CGI in DMVs, CGI in all promoters, and non-CGI in all promoters). The scale of the circles represents motif enrichment. Red represents homeobox TFs. e Venn diagram showing the overlap between DMVs and hyper-conserved CpG island domains (HCGDs) [41] in human. f Boxplot showing deamination rates [42] for human CGIs inside or outside DMVs classified based on their CG densities