Fig. 1.

EpiCHAOS reliably quantifies epigenetic heterogeneity in single-cell epigenomics data. A Schematic describing epiCHAOS calculation. Using single-cell epigenomics data in binarized matrix form, epiCHAOS scores are assigned per cluster by computing the mean of all pairwise cell-to-cell distances using a chance-centered Jaccard index followed by regression-based adjustment for sparsity. μ = mean per cluster. B Scatter plot illustrating the correlation between epiCHAOS scores (epiCHAOS) and controlled heterogeneity across 100 synthetic datasets. Pearson correlation coefficient and p-value are shown. C Barplots illustrate increasing heterogeneity after perturbation of scATAC-seq data from sorted monocytes by either randomly adding or randomly removing 10–50% of 1’s. D Boxplot comparing epiCHAOS scores across six simulated single-cell ATAC-seq datasets with varying sequencing depths. Data were simulated using scReadSim with sequencing depth varying from 50,000 to 100,000 counts. ScATAC-seq data from the hematopoietic stem cells subset from the Granja et al. dataset [28] were used as the baseline counts matrix. E Validation of epiCHAOS using in silico mixtures of hematopoietic cell types. UMAP embedding illustrates scATAC profiles from five selected cell types of human bone marrow [28]. After selecting 500 top differentially accessible peaks for each cell type, in silico mixtures of two to five cell types in all possible combinations were created. Boxplots show the relationship between epiCHAOS scores (epiCHAOS) and the number of cell types (y-axis) after in silico mixing