DNA methylation and gene expression capture complementary information. (A) Two-dimensional hierarchical clustering of genes differentially expressed between the 2 leukemia subgroups, illustrated by a heatmap. Supervised analysis identified 587 probe sets (415 genes) at a P < .001 and fold change more than 2. Cases are represented in the columns; probe sets are represented in the rows. CEBPAsil cases are clustered in the right node, and CEBPAmut cases are clustered in the left node. (B) Density (y-axis) plot for the gene-by-gene correlations (x-axis) between gene expression log intensity and log(HpaII/MspI) values. A positive correlation between these 2 measures translates into a negative biologic correlation, that is, hypermethylation in combination with lower expression levels, or hypomethylation in combination with higher expression levels. In black, density plot for the correlation between expression and methylation for a set of 600 randomly selected probe sets. In red, density plot for the correlation between expression and methylation measured by the 567 probe sets in the genes differentially methylated between the 2 subgroups. The shifting of the density plot to the right reflects a tendency to a stronger correlation of DNA methylation with gene expression levels in this subset of genes. The figure is representative of 5 analyses, each time using a different set of 600 randomly selected HpaII-amplifiable fragments for the calculation of correlations.
