Fig. 4
From: Controlling for conservation in genome-wide DNA methylation studies
The COMPARE and Paired-Region Averaging methods for correcting biases introduced by differences in conservation rates. Top: COMPARE learns a different set of parameters for every column, and the methylation tendency of site i,j is inferred from the features of row i and the parameters set for column j. The complete inferred matrix can then be used in subsequent average-based tests. Bottom: Paired-Region Averaging computes for each instance (e.g. a specific intron-exon junction) the average methylation rate at each side of the functional boundary and excludes instances at which either of the boundary sides is lacking any values (and hence no average-based DNA methylation rate can be assigned). Tests to compare the relative methylation tendencies of the different boundary sides can be applied, but a quantitative analysis of the DNA methylation rates cannot be conducted when using this method due to the region-exclusion step