Fig. 1

Gene co-regulation in chromatin domains. a–c Schematic of the DADo algorithm: a DADo integrates Hi-C (top: example of Hi-C contact map) and gene expression data (bottom: toy barplot showing concordant expression) to assess when genes (e.g., 1, 2, and 3) within a chromatin domain (black triangle in the Hi-C map) exhibit coordinated expression differences between 2 conditions (e.g., A and B). b First, the fold-change concordance (FCC) score is computed and the ratio between the observed (red) and expected (gray) areas under the curve (AUC) are computed. c Next, the mean gene expression fold-change (FC) and mean mRNA correlation among all genes in the domain are computed and used to determine differentially active domains. d Heatmap representation of the density distribution of FCC scores (range: −1 to 1; X-axis) for each of the 58 comparisons (Y-axis) made by matching 30 Hi-C and 12 mRNA expression datasets (left). Barplot of the FCC AUC ratios for all comparison (right). Asterisk (*) indicates the top ranking dataset (lung tissue—lung adenocarcinoma (LUAD)—EGFR-mutant vs. KRAS-mutant—see panel e). e Example of the FCC distribution (left) and the cumulative sum curves (right) for the dataset with highest FCC AUC ratio (lung tissue—LUAD—EGFR-mutant vs. KRAS-mutant). f For each dataset (X-axis), the percentage of chromatin domains that are fully concordant (i.e., FCC = 1) is shown for the real data (red dots) and the randomized data (gray dots). g Ratio of fully concordant domains comprising only 3 genes for the observed (red, left) and average of permutation data (gray, right).