Fig. 2: Bias in negative coefficients from regression analysis indicates regionally specific DNA hypomethylation profile in ALL cases. | Nature Communications

Fig. 2: Bias in negative coefficients from regression analysis indicates regionally specific DNA hypomethylation profile in ALL cases.

From: DNA methylation at birth in monozygotic twins discordant for pediatric acute lymphoblastic leukemia

Fig. 2

a Box and whisker plots demonstrate median DNA methylation content in n = 41 case and n = 41 control twins (pairs connected by gray lines). Significantly lower global DNA methylation was identified in the case of twins when compared to their sibling controls (Two-sided paired Wilcoxon test P = 0.048). b Bar plot demonstrating proportion of positive (indicating hypermethylation in cases) and negative (indicating hypomethylation) regression coefficients in all probes, by CpG island context, and by genetic context. The percentage of positive and negative probes are shown over bars. All regions are significantly biased toward negative regression coefficients. Regions are arranged by the strength of negative coefficient bias, with CpGs in 3′-UTR, gene body, open sea, exon boundaries (Exon Bnd), and CpG island shelf/shore regions showing a stronger negative bias than the overall array results. ** indicates FDR-corrected P < 0.001. Evaluation of DNA hypomethylation signature identified in conditional regression analysis using raw beta values paired by twin relationship (delta beta, or ALL-case DNA methylation beta value minus control beta value) according to c CpG context and d UCSC RefGene Group for n = 41 independent twin pairs. *FDR-corrected P < 0.05. The distribution of median delta beta values (case beta minus control beta) for probes associated with each genomic region is shown for the full set of 41 twin pairs. A significant shift toward DNA hypomethylation in cases (negative delta beta values) was identified for probes associated with the overall array (All), as well as open sea and shelf/shore regions, however not in island regions. When looking at pair-specific median values (e), twin pairs with negative median delta beta values across all probes (globally hypomethylated in cases, n = 30) tended to have median island values near zero, demonstrating regional specificity of the identified DNA hypomethylation profile. Plot colors represent individual twin pairs. Similarly, by RefGene group (f), a significant negative shift was noted in the gene body, 3′-UTR, and 5′-UTR, at exon boundaries, and at TSS1500 sites, however, there was no negative shift associated with 1st exon and TSS200 probes. Among globally hypomethylated and hypermethylated cases, a pair-specific plot (g, h) shows specificity of findings between gene body and promoter-associated probes. Source data are provided as a Source data file. In boxplots, the box represents the interquartile range (IQR, first through the third quartile) with the centerline showing the median value for all subjects/twin pairs, whiskers show a minimum (first quartile minus 1.5 × IQR) and maximum (third quartile plus 1.5 × IQR) data range.

Back to article page