Figure 5.

DMGs are enriched in distinct pathways and may contribute to prognostic differences between clusters. (A) Gene enrichment for DMRs (y-axis = −log10 [P value] from hypergeometric test of enrichment for DMRs present within a gene ±1 kb, relative to all CpG tiles within a gene; x-axis, mean methylation difference of DMRs within gene ±1 kb). (B) Heat map of average methylation for the most highly DMGs for each set of cluster-specific DMRs (top: samples in columns, genes in rows). WT1 was the most highly DMG for both clusters B and E. Heat map of the average methylation of genes in significantly enriched biological process gene ontologies from GO enrichment analysis of cluster-specific DMGs (bottom: samples in columns, genes in rows; biological process ontology indicated by left panel of heat map). (C) Methylation of cluster-specific differentially methylated CpG tiles within the WT1 gene (second panel). Gray points represent the average methylation of patients in clusters A, C, and D, which all had similar levels of methylation at these loci. Reduced representation bisulfite sequencing (RRBS) methylation as well as H3K27ac, H3K4me3, and H3K4me1 Chip-seq signal tracks correspond to a reference dataset of granulocyte colony-stimulating factor mobilized CD34⁺ cells from a healthy male donor (third through sixth panels; Roadmap Consortium epigenome E051). Gene model track (top panel) corresponds to isoforms of WT1 found in RefSeq with their corresponding Ensembl IDs. (D) Kaplan-Meier curves for the Gerstung et al 2015 cohort stratified by average WT1 expression (left) and for the present study stratified by average WT1 regulatory region methylation (region ± 2500 bp from TSS) (right).