Clustering analysis. 1a. Hierarchical euclidean based clustering of the 2000 most variable loci. Samples split into 2 major cluster groups designated as being either CIMP+ve or CIMP-ve with each major cluster splitting into 2 sub-groups. Normal samples clustered together and are labeled N1-N4, tumor samples are labeled with their pair number (P#) followed by either E or L to denote early or late lesion respectively. 1b. Box and whisker plots of cluster group ANOVAs. 1c. CIMP status as determined by clustering is shown with a black or white circle representing CIMP+ve and CIMP-ve respectively. IDH1 mutation status is shown as either mutant (mut) or wild-type (wt) for the p.R132H change.