Fig. 1
Discovery and characterization of differential methylation characteristic of short term survivors. a Boxplot showing the global distribution of methylation for LTS (n = 22) and STS (n = 14) samples in cohort 1. Y-axis—mean methylation β value for each sample. Each dot represents the mean methylation per sample. The non-parametric Wilcoxon test P value shows no significant difference between groups, although there is a trend toward hypermethylation in STS tumors. b Volcano plot showing differential STS/LTS methylation. X-axis: mean change in methylation (Δβ(STS—LTS)) for each CpG, y-axis: − log10 (P value) using a t test. The dotted horizontal line represents the P value cutoff of 0.01; two dotted vertical lines delineate the Δβ(STS–LTS) cutoff set at − 0.1 and 0.1. Each dot represents a CpG: sites in red (n = 5929) meet the P value and Δβ cutoffs, and are thus referred to as differentially methylated CpGs (DMCpGs). c PCA plot using the 5929 DMCpG shows separation of pathologic groups into distinct clusters. LTS-red dots, STS-blue dots. The percentage variation between groups explained by each of the principal components is indicated. d Heatmap showing supervised hierarchical clustering of LTS and STS using the 5929 DMCpGs reveals two distinct clusters marked by red rectangles: one is hypermethylated (n = 4570) and the other hypomethylated (n = 1359) in STS. Color bars underneath the column dendrogram represent, from top to bottom: sex, SSIGN score, and disease status. The color bar next to the row dendrogram indicates genomic feature. e Barplot showing the relative distribution of 4570 hypermethylated and 1359 hypomethylated DMCpGs normalized to the distribution of all 2.4 M CpGs in cohort 1 over four genomic features (intergenic, enhancer, promoter, and body). The distribution of all features is significantly different between the sites in cohort 1 and the total DMCpGs. Y-axis—fold change (log10) of each feature. f Venn diagram showing the number of DMCpGs overlapping a selection of three histone marks characteristic of regulatory regions: H3K27ac, H3K4me1, and H3K4me3. N = 698 are not represented as they do not overlap with any of the three marks