Figure 5.

Variation filtering mostly, but not completely, removes quantification bias. Filtering out strain-unique CpGs from the BL6 reference sufficiently addresses global methylation bias (A). However, an example 11 kb genomic region, despite containing no CpGs unique to either mouse strains, appears as a DMR under the filtering strategy (B) despite no true methylation difference under personal-genome alignment (C). This is due to misalignment of reads: upwards tick marks indicate CpGs with average coverage >100 in the filtered analysis.