Coordinates for Refseq gene annotations (rn4), CpG islands (‘AL' UCSC track)(Gardiner-Garden and Frommer, 1987), and common repeats (RepeatMasker; see Supplementary Materials and Materials and Methods for more details) (Jurka, 2000; Smit et al, 1996–2010) were downloaded from the UCSC genome browser (http://genome.ucsc.edu/) (Kent et al., 2002). The overlap of CpGs with annotated genomic features were assessed using BedTools version 2.1 (Quinlan and Hall, 2011). Refseq gene annotations were segregated into four categories: exonic (within coding or 3′UTR sequence), intronic, promoter (±2 kbp of transcription start site (TSS)) and intergenic (outside of exons, introns, and promoters). Likewise, CpG island annotations were classified as one of four categories: CpG islands (CpGi); CpGi shores (±2 kbp from CpGi); CpGi shelves (±2 kbp from CpGi shores); and sea (outside of all other CpGi annotations). In cases where CpGs fell within more than one CpGi annotation, islands took precedence over shores and shelves, and shores took precedence over shelves. For both gene and CpGi feature overlaps, the distribution of DMR-CpGs was assessed using the χ2-test to compare against the background list of CpGs that included all tested sites (n=567 306; referred to as ‘All' below).
