Fig. 2

Gains of ASM in cancers due to widespread allele-specific LOM. a Schematic showing the average configurations for allelic methylation levels in non-cancer and cancer samples at loci where ASM was observed only in cancer. Cancer samples are compared to lineage-matched non-cancer cell types. Average fractional methylation was estimated using a linear mixed model with random intercept and random slope (the “Materials and methods” section). For each sample type, the squares represent the model estimate of the average fractional methylation in the low and high methylated alleles. b Examples showing primary WGBS data. For the three types of cancers, the most frequent situation is an allele-specific LOM occurring in the cancers at loci that are highly methylated in the lineage-matched normal cell types. Methyl-seq reads separated by REF and ALT allele. Methylated CpGs are black and unmethylated CpGs are white. c Distribution of net methylation levels in normal B cells (left) and glia (right) grouped into 3 classes (low, intermediate and high methylation) at all informative CpGs (random expectation) and at CpGs where ASM was observed in the cancers but not in the matched non-cancer samples. While allele-specific LOM in cancer accounts for most instances of cancer-only ASM (black bars; high methylation in the normal samples), relative to the background of global hypomethylation in the cancers, it occurs less often than random expectation. In contrast, the smaller group of loci that have GOM leading to cancer-only ASM (white bars; low methylation in the normal samples) show a significant enrichment over random expectation, given the globally hypomethylated genomic background of the cancers