Research Paper Volume 12, Issue 21 pp 21037—21056

DNA methylation-based age estimation in pediatric healthy tissues and brain tumors

Figure 5. Subtype specific age acceleration in pediatric brain tumors assessed by Horvath methylation clock. (A) Age acceleration (Horvath methylation age minus chronological age) varies significantly (adj. p < 0.05) between all three ATRT subtypes in GSE90496 (n = 91; left panel), and the validation cohort (n = 49; right panel) shows the same trend with a significant difference (adj. p = 0.023) between the TYR and SHH subgroup. (B) In ependymoma (EPN), the YAP subgroup displays significantly lower age acceleration than the RELA (adj. p = 1.3e-6) and PF-A (adj. p = 1.0e-7) subgroups in GSE90496 (n = 157; left panel). Right panel with validation cohort (n = 65) shows significant difference between the RELA and PF_A subgroup (p = 0.02). (C) Left panel, GSE90496 (n = 295), is significantly different (adj. p < 0.05) for all pairs of gliomas except PXA (pleomorphic xanthoastrocytoma) vs LGG (low-grade glioma, mainly grade I) and K27 (diffuse midline glioma H3 K27M mutant), and GBM (glioblastoma) vs G34 (glioblastoma H3.3 G34 mutant). Middle panel with validation cohort (n = 153) varies significantly (adj. p < 0.05) between all pairs except K27 vs GBM and G34 and G34 vs GBM and DIPG, and right panel with local cohort (n = 86, p-value = 0.003). (D) The medulloblastoma (MB) subtypes in GSE90496 (n = 306; left panel) are significantly different for all pairs (adj. p < 1e-8) except SHH vs G4 (Group4). The age acceleration in the validation cohort (n = 48; middle panel) displays a similar trend in decreasing age acceleration, and in the local cohort (n=39; right panel) there is a significant difference between all pairs (adj. p < 0.05). Note that the subtypes in A-D have been ordered according to prognosis where the left-most subtype have the best prognosis and the right-most have the worst. Sample sizes for all included boxes is available in Supplementary Table 3.