Extended Data Fig. 6: Characterization of scEpiChem data in human CRC organoids.
From: Single-cell EpiChem jointly measures drug–chromatin binding and multimodal epigenome
a Spearman correlation of THZ1-btn and Dox-btn binding signals in bulk and single cell data within the TSS 5±kb regions in human CRC organoids. b ROC curves for CRC bulk and single cell aggregated THZ1-btn (4,424 cells) or Dox-btn (4,793 cells), using CRC bulk THZ1-btn or Dox-btn data as gold standard. AUC: THZ1-btn, 0.9561; Dox-btn, 0.8361. c UMAP showing scEpiChem data (small molecules and H3K27ac) on CRC organoids (n = 14,027 cells) with projections of the gene activity scores of the EMT score and CDH1. The EMT score was calculated based on 1,184 signature genes (http://www.dbemt.bioinfo-minzhao.org/). d UMAP showing the batch effect of scEpiChem (small molecules and ATAC; small molecules, target and ATAC) on CRC organoids data from two samples, p1201 (n = 6,209) and p20 (n = 3,222); p1201 (n = 12,790) and p20 (n = 8,993). e Track view of H3K27ac and Dox-btn signals on the representative loci in CRC organoid cells. f Genomic distributions of scEpiChem peaks of Dox-btn and H3K27ac in CRC organoid cells. g Venn diagram showing peak overlap of Dox-btn and H3K27ac in CRC organoid cells. h GO terms enriched by Dox-btn-specific peaks. P-value was calculated by binomial test. The results of GO term enrichment analysis using a hypergeometric test, with two-sided statistical tests and adjustments for multiple comparisons employing the Benjamini-Hochberg method. i-j Distribution of the median non-duplicated reads for single cells in H3K27ac (n = 12,066), Dox-btn (n = 3,052), JQ1-btn (n = 4,012) and THZ1-btn (n = 3,453) with scEpiChem (small molecules and H3K27ac) (i) and ATAC (n = 10,672), BRD4 (n = 5,809) and JQ1-btn (n = 5,302) with scEpiChem (small molecules, target and ATAC) (j) in CRC organoids data. Boxes in the violin plots: center marks the median; edges of boxes define the 25th and 75th percentiles.
