Fig. 3

Analysis of DNA methylation at functional boundaries. Matrices containing bisulfite-generated methylation estimates were constructed from the relevant instances of the human genome (hg18), incorporating 4852 5′UTR-coding junctions, 20,784 mid-gene intron-exon junctions and 21,408 CG island junctions. Each plot shows average methylation measurements at increasing distances relative to the junction (different columns of the matrix). The means of the column-average values are shown as dashed lines. Simulation (leftmost column)- the methylation values for each pair of coupled matrix rows were shuffled according to a random permutation (maintaining the locations of missing values), resulting in simulated data with equal methylation tendency across the junction. Differences observed in the mCG/CG measures are due to differences in the matrix structure across the boundary (as in Fig. 1). %mCG Analysis (middle)–naïve averaging of the values in the matrix columns. Corrected (rightmost columns)–column-averages after inference with COMPARE and comparisons after correction with Paired Region Averaging