Figure 6. Human CpG islands show evidence of selection at methylation signature motifs.

The distribution of predicted DNMT3B methylation coldspots (5′-NGCGCN-3′, top) and hotspots (5′-NTCGGN-3′, bottom) was analyzed in two datasets corresponding to all “specific” and all “weak” CpG islands from human chromosome 1, respectively. The score on the y-axis represents the exceptionality of the frequency of a given motif. A positive score indicates an over-representation of a motif in a given model of the sequence (the M1 model, which computes expected motifs frequencies based on mono- and di-nucleotide frequencies was used here). A negative score indicates under-representation. A higher absolute score value indicates a greater statistical significance (a score of 5 corresponds to a p-value of ∼10⁻⁶, which starts to be considered significant; a score of 10 corresponds to a p-value of ∼10⁻²³).