Extended Data Fig. 4: Comparison of CG and CH DMRs across studies.
From: Transient naive reprogramming corrects hiPS cells functionally and epigenetically
a) Upset plot shows number of CG-DMRs detected for this study and how they overlap with CG-DMRs detected from previously published data processed using identical methods. b) Difference in DNA methylation level between hiPSCs and hESCs at CG-DMRs identified between Primed-hiPSCs and hESCs. Vertical dashed lines indicate the threshold of 20% minimum difference in CG DNA methylation level at CG-DMRs. c) Enrichment z-score determined from permutation testing of enrichment of CG-DMRs in repressive chromatin domains and of d) CH-DMRs in published studies. e) Heatmap of CA methylation levels in CH-DMRs in this study and previously published studies showing Primed-hiPSCs from all studies clustering separately to hESCs. f) Genome track of a CH-DMR region that intersects a PMD, fibroblast lamina associated domain (LAD), and clusters of CG-DMRs in each study. g) Principal component analysis of CG methylation levels in CG-DMRs for all studies combined. Top left plot shows the proportion of variance explained by each principal component. Scatter plots with coloured points show principal component separation of hESCs, Primed-hiPSCs, and TNT-hiPSCs. Ellipses around points indicate 95% confidence interval for a multivariate t-distribution. These data indicate that principal component 3 (PC3) in the bottom left plot clearly separates Primed-hiPSCs and hESCs for all studies, and shows that TNT-hiPSCs are more similar to hESCs by this measure. h) Plots of eigenvalues for each principal component for Primed-hiPSCs, TNT-hiPSCs, and hESCs, and i) data split by study/lab. Red bars indicate P < 0.05 for one-way ANOVA, with FDR reported above red bars.
