Hierarchic cluster analysis of AML samples and their methylation profiles. Two-way hierarchic cluster analysis of 182 acute myeloid leukemia (AML) samples (rows) and DNA methylation of CpG units in 92 promoter regions (columns). DNA methylation values are depicted by a pseudocolor scale as indicated (methylation increases from red [nonmethylated] to white [methylated]). Samples with overall poor data quality were removed before clustering. Gray denotes data of poor quality. Samples are color-coded according to the underlying cytogenetic aberration (legend depicted top left). Most samples with an inv(16) were located in one cluster (highlighted in yellow); most t(11q23) cases grouped together (highlighted in green) and translocation t(15;17) and t(8;21) cases did also cluster mainly in one group (highlighted in pink). A small set of 3 samples splits off early in the dendrogram showing generally hypermethylated DNA (bottom 3 rows, highlighted in blue). CpG units form 2 large clusters that are characterized mainly by the level of methylation. The larger group (group 1, highlighted in pink) is characterized mainly by unmethylated CpG units, the smaller one (group 2, highlighted in yellow) mainly by methylated CpG units. In each cluster a group of CpG units can be separated that shows more variable methylation values across the AML samples (as indicated). The black bars on top of the heatmap denote the top 10% of CpG units with the highest variance across the AML samples (> 90th percentile).