Extended Data Fig. 2: Glioma Methylation Signature identification.
a, DMRs on chromosomal region chr6p22.2-21.33 and overlapping genes of top DMRs (P < 1×10−40, n = 37). b, DMRs on chromosomal region chr11q12.2-13.5 and overlapping genes of top DMRs (P < 1×10−21, n = 41). c, GO enrichment analysis of genes overlapping with on chr11q12.2-13.5. Top 10 pathways are shown, the -log10(Pvalues) for overrepresentation of the gene-set are obtained using Wallenius’ noncentral hypergeometric distribution and FDR adjusted using BH correction. Xaxis depicts the percentage of genes from the gene-set that were differentially methylated, GO subontologies are indicated at the end of the bar for each gene-set. d, Regression coefficients plot of multiple linear regression models testing the effect and pair-wise interaction of known molecular markers of poor prognosis to the GMS-score (GMS ~ (markers)^2) in matched initial (circles) and recurrent (squares) tumor samples (n = 103). Estimates and 95% CI are plotted, the n for each molecular marker is indicated, p-values were calculated using a two-sided Wald test. e, Histogram depicting the distribution of GMS-scores of all GLASS-NL DNA-methylation samples. Bars are colored based on the assigned GMS-class (‘hyper’: >1 M; ‘intermediate’: 0 – 1 M; ‘hypo’: <0 M). f, Kaplan-Meier curve for OS when stratifying patients on the GMS-class of the initial tumor (‘hypo’ vs ‘hyper’, P = 0.699, HR = 1.326 (0.3162-5.564); ‘intermediate’ vs ‘hyper’: P = 0.105, HR = 1.929 (0.8720-4.266), two-sided Wald test). g,h, Multivariate CoxPH models on TTR and PRS with patient and tumor sample characteristics related to initial and recurrent surgery, respectively, including the M-value of all CpGs (DNAm Score) (g) or the GMS-score (h) as a continuous variable. Hazard ratios and 95% CI are plotted, P values are indicated. i, Kaplan-Meier curve for PRS when stratifying patients on the GMS-class and WHO grade of the recurrent tumor. WHO grade 2 and 3, and GMS-class hyper and intermediate are grouped together. Compared to the ‘WHO23/GMShigh’ group, all other groups of recurrent tumors were negatively associated with PRS (‘WHO23/GMSlow’, P = 0.004, HR = 6.306 (1.804-22.042); ‘WHO4/GMShigh’, P = 5.6×10−4, HR = 3.462 (1.711-7.005); ‘WHO4/GMSlow’: P = 1.35×10−10, HR = 8.031 (4.252-15.167), two-sided Wald test). j, Volcano plot depicting the CpG-level temporal changes in DNA-methylation of patients with CNS WHO grade 4 initial and recurrent tumors (n = 10). Differentially methylated CpGs, from a top 10.000 CpG selection, were identified using a paired two-sided Wilcoxon signed-rank test (BH-corrected, FDR < 0.05 & ΔM > 1).
