Figure 2

Hierarchical unsupervised clustering of 31 study samples. DNA methylation levels at the 53 most significant CpG sites were used for hierarchical unsupervised clustering. The results show a clear distinction between the 10 KDM5C mutation cases (top part), and the remaining 21 samples (bottom part), which is also indicated by the large distance between the two branches of the dendrogram on the left. DNA methylation levels corresponding to mutations are generally lower than for the controls. Interestingly, the two cases of variant of unknown clinical significance (p.R1546Q) have methylation profiles very similar to the 19 controls. Also, mutation cases from the same family do not always cluster together. The clustering was based on complete linkage and cosine distance metric.