Figure 3

TCGA replication analysis. HM450 genome-wide methylation data was downloaded from TCGA (LIHC) for 47 available HCC tumor/surrounding pairs (see Methods). (A) Multi-dimensional scaling (MDS) plot showing differential clustering of control vs. tumor tissues. (B) Differential methylation analysis using paired (tumor/surrounding) linear regression identified 1,328 differentially methylated positions (DMPs); 30 sites (out of 1,328) corresponding to known imprinted loci were used to build an unsupervised cluster. The corresponding heatmap shows the normalized methylation data in a blue-red scale (from lower to higher methylation). Tumor grade is shown in the annotation panel as G1, G2, or G3. (C) Example plots of methylation data for 6 of the top differentially methylated imprinted sites in the TCGA HCC tumor vs. surrounding comparison (n = 47 tumor/surrounding pairs). (D) Enrichment analysis comparing the proportion of differentially methylated positions (DMPs) found in imprinted loci for dataset 1 (our original GoldenGate analysis) and dataset 2 (TCGA HM450 analysis). The barplot shows a significant enrichment in imprinted loci in both datasets. The data used to calculate enrichment and the corresponding P values (Fisher’s exact test) are shown in the lower panel. *P value <0.05.