Fig. 3: epi-gSCAR performance and validation.
From: Bisulfite-free epigenomics and genomics of single cells through methylation-sensitive restriction
a Global single-cell methylation levels of Kasumi-1 (red dots) and OCI-AML3 (blue dots), and corresponding mean methylation levels (black lines) in comparison to mean methylation levels of WGBS (violet line) and 450 K array (green line) from cell-bulk samples. b Pairwise CpG concordance for all analyzed Kasumi-1 single cells. c Pairwise CpG concordance for all analyzed OCI-AML3 single cells. CpG concordance was calculated for all overlapping CpGs between each single-cell pair of each cell line as the fraction of CpGs with the identical methylation state (0, 0.5, or 1). Calculations are based on 66,588–297,161 CpGs for Kasumi-1 single-cell pairs and 148,932–257,837 CpGs for OCI-AML3 single-cell pairs. d Correlation between the mean pseudo-bulk methylation and the cell-bulk 450 K array, and WGBS datasets for Kasumi-1 and OCI-AML3. Comparisons consider genome-wide methylation of individual CpGs covered in ≥15 single cells (Kasumi-1 450 K: n = 6,607; Kasumi-1 WGBS: n = 20,000 of 72,142 CpGs covered by epi-gSCAR in ≥15 single cells; OCI-AML3 450 K: n = 11,511; OCI-AML3 WGBS: n = 20,000 of 129,153 CpGs covered by epi-gSCAR in ≥15 single cells). e Circos plot representation of genome-wide methylation profiles of randomly selected single cells, the pseudo-bulk datasets, and WGBS controls. The heatmaps show average methylation levels for 200 kb windows. Heatmap colors indicate methylation levels from low (blue) to high (red). Tracks from inside to outside represent single cells O_01, O_05, O_11, K_11, K_16, and K_17, OCI-AML3 pseudo-bulk (O_01–O_20), OCI-AML3 cell-bulk WGBS, Kasumi-1 pseudo-bulk (K_08–K_27), and Kasumi-1 cell-bulk WGBS. f Hierarchical clustering analysis based on Pearson correlation coefficients for single cells K_01–K_07 (yellow), K_08–K_27 (red), and O_01–O_20 (blue) across 200 kb windows. g Multidimensional scaling analysis using UMAP, in which each dot represents a single cell (K_01–K_07, yellow; K_08–K_27, red; and O_01–O_20, blue). Cells are clustered based on the methylation levels across 200 kb windows covered in all single cells (top; n = 10,555) or based on genetic variants called at positions covered in all single cells (bottom; n = 7,027). For genetic variant clustering, SNV data was converted into a categorical numeric matrix as an input to compute UMAP with the R package ggplot.
