Fig. 1.

Array design bias for the Illumina HumanMethylation EPIC BeadChip. A Frequency plot of the numbers of CpGs measuring methylation across each gene for the EPIC array (probe number bias). The most extreme value is 1485 CpGs measuring methylation across a single gene. The median is 20 and the first mode is 4. B Plot demonstrating probe-number bias for B cells vs. NK cells from sorted blood cell type EPIC array data. Genes with more measured CpGs are more likely to be differentially methylated. C Histogram of the median numbers of CpGs per gene for each GO category for the 450K and EPIC arrays. The distributions differ between the arrays; however, both show a varying number of CpGs per gene per GO category. GO categories with more CpGs per gene, on average, have greater power to be significantly enriched. D Split bar chart showing the numbers of genes annotated to each CpG (multi-gene bias). While the majority of CpGs are annotated to only one gene, there is still a large number annotated to 2 or more genes