Fig. 2

Overview of how probe number and multi-gene bias is taken into account in GOmeth. CpGs are not evenly spaced throughout the genome. Gene 1 has methylation measured at three CpGs, and gene 2 has methylation measured at eight CpGs. The CpG shaded in blue is an example of a shared genomic location between genes 1 and 2, with one CpG measuring the methylation status for two genes, and is thus not independently measured. By calculating a weight for each CpG inversely proportional to how many genes that CpG is annotated to, we can calculate the equivalent numbers of CpGs measured across each gene and ensure the enrichment statistic for Wallenius’ noncentral hypergeometric test is not artificially inflated due to multi-gene annotated CpGs