Fig. 3: Distal peaks lend cell-type uniqueness and reveal underlying cell-type-specific pathways and phenotypes.

a Heatmap of differential cluster-specific peak accessibility (FDR ≤  0.01, log2-fold change ≥ 0.5) parsed into distal (n = 27,295 peaks) and promoter (n = 2436 peaks) with clusters ordered by unsupervised hierarchical clustering on the distal peaks and row z-scored. TEP, thymic epithelial progenitor. b Boxplot showing cluster specificity of differentially accessible distal (n = 27,295) and promoter proximal (n = 2436) peaks measured by cluster-specificity index on scaled, log2-transformed cluster averaged peak accessibility. Center line denotes the median; box limits denote upper and lower quartiles; top and bottom whiskers denote regions upto 1.5 times the inter-quartile range above the upper and below the lower quartiles, respectively; tailing points denote outliers. P-value is from a two-sided Wilcoxon test with continuity correction and without adjustment. c Heatmap of top 10% of distal regulators per cluster determined by a motif score (average ChromVar motif z-score per cluster – minimum cluster average motif z-score) on a set of 55,840 differentially accessible distal peaks (FDR ≤ 0.1, log2-fold change ≥ 0.5) with color bar displaying correlation of the ChromVar motif z-scores with the integrated gene expression (correlation cutoff > 0.35). d Percent of distal differentially accessible peaks between clusters (FDR ≤ 0.01, log2-fold change ≥ 0.5) having an average phastCon conservation score within the Euarchontoglires clade = 0.00, >0.00 and ≤0.50, and >0.50 split by cluster. Cluster 20 has no peaks meeting the differential accessibility cutoff. e Heatmap of row normalized −log10 corrected p-values (Benjamini–Hochberg) from the binomial test on a curated list of GREAT terms enriched in distal, differentially accessible peaks (FDR ≤ 0.01, log2-fold change ≥ 0.5) with an average phastCon conservation score within the Euarchontoglires clade of >0.50.