DNA sequence can predict methylation levels in PMDs in IMR90.

<p>(a) Performance of the positional dinucleotide model (see <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1004143#s4" target="_blank">Materials and Methods</a>) in PMDs. Depicted is the 2D-density for a total of 100K CpGs in PMDs. The x-axis denotes the measured methylation level and the y-axis denotes the predicted methylation level. (b) Inferred parameters of the positional dinucleotide model (see <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1004143#s4" target="_blank">Materials and Methods</a>) which predicts methylation levels in PMDs. The contributions of the 16 dinucleotides within a window of +−60 bp around the central CpG are color coded and displayed as a heatmap. (c) Same as (b) for the indicated tissues, separately for PMDs and regions outside of PMDs. For H1, which does not contain PMDs, the PMD annotation of IMR90 was used. In contrast to (b), CpGs overlapping DHS were not removed as the necessary DHS dataset were not available for all cell types. Thus panel b) and the upper left panel in c) only differ with respect to the inclusion or exclusion, respectively, of CpGs based on DHS. (d) Performance of the dinucleotide model for the same tissues and regions as in (c). (e) Contribution of each dinucleotide to allelic methylation differences in PMDs. The sketch at the top depicts the investigated genomic configuration, namely a CpG conserved in both alleles but flanked by a SNP. The difference of one nucleotide is used to assign the bisulfite reads covering the SNP to an allele (Reference or Alternative). The reads that cover both, the SNP as well as the conserved CpG are used to quantify the allele specific methylation level of the conserved CpG. Given the nucleotides involved in the SNP and the difference in methylation level between the two alleles, we calculated the allele specific contribution of each dinucleotide at a given distance from the conserved CpG (see <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1004143#s4" target="_blank">Materials and Methods</a>). The analysis was performed within a window of +−40 bp due the low number of data points available at larger distances.