Supplementary Figure 1: Analysis of DNA methylation in a cancer cohort based on Infinium 450K data.

RnBeads was used to rediscover a clinically distinct subgroup of glioblastoma patients characterized by increased DNA methylation levels (termed G-CIMP+), and to predict the G-CIMP status for a total of 124 patients using Infinium 450k data obtained from the TCGA project (http://cancergenome.nih.gov).

(a) Detection of genetic duplicates among the patient samples (columns) using a clustered heatmap of intensity values for the genotyping probes that are present on the Infinium microarray (rows). The inset shows that two samples exhibit a high level of genetic identity, and they are indeed derived from tumors of the same patient.

(b) Quality control plot summarizing the outcome of the data filtering. The bar plots on the top left show that the majority of CpG sites (top) and samples (bottom) are of good quality and can be retained. The relatively straight line in the quantile-quantile plot indicates that the probe filtering does not have a major impact on the distribution of DNA methylation in the dataset.

(c) Identification of a small but clearly distinguished cluster of G-CIMP+ glioblastoma samples with elevated DNA methylation levels especially in CpG-rich genomic regions (light blue in the leftmost column). In the heatmap, blue colors denote high levels of DNA methylation, red indicates low levels and grey represents intermediate levels. For visualization purposes, only the 100 gene promoters (rows) with the highest levels of inter-sample variation in DNA methylation are shown (columns), but the hierarchical clustering is based on the full set of promoters.

(d) Global assessment of the similarity between the DNA methylation profiles, plotting all glioblastoma samples according to their second and third principal components. The samples exhibit strong separation according to the G-CIMP status (denoted by point shape) and IDH1 mutation status (denoted by point color).

(e) Analysis of significant associations between all user-provided sample annotations. Significant p-values (<0.05) are highlighted in the left triangle, and the corresponding statistical tests are annotated in the right triangle (orange: Pearson correlation followed by permutation-based estimation of the p-value; green: Fisher’s exact test; blue: Wilcoxon rank sum test; violet: Kruskal-Wallis one-way analysis of variance).

(f) Genome-scale comparison between the DNA methylation levels of G-CIMP positive (y-axis) and G-CIMP negative (x-axis) tumor samples, focusing on CpG islands (left scatterplot) and on 5-kilobase tiling regions with a CpG content in the bottom quartile (right scatterplot), respectively. Genomic regions that are differentially methylated with an FDR below 0.05 are presented as red points. All other regions are displayed in blue, and color brightness denotes point density.