Extended Data Fig. 1: 3D chromatin structure modeling and assessment.

a, The contact probability matrix calculated from Hi-C (left) and the structure population (right) for chromosome 2. Zoomed-in heatmaps show the matrix for 40–80 Mb. b, Density scatter plot of contact probabilities from Hi-C data and structure population (Pearson correlation: r = 0.98, p = ~0). c, Histograms of restraint residuals from the structure population (Methods). A restraint residual less than 1.05 is considered satisfied and is not displayed in the histograms (99.9% of restraints fall in this category). d, The contact probability matrix for chromosome 2 showing the 50% randomly chosen dataset used as input (lower triangle) vs. the matrix generated from the structure population (upper triangle). e, Density plot of Hi-C contact probabilities missing in the input and their predictions from the structure population (Pearson correlation: r = 0.93, p = ~0). f, Average radial positions of chromatin in different replication phases⁵⁸. Numbers of regions in each boxplot are (from left to right): 1,714, 1,921, 2,166, 2,246, 2,068, 2,064. g, Comparison of the inter-chromosomal loci co-localization frequencies between the observed occurrence in FISH experiments¹⁸ and in the structure population shown as bar plots (left), and scatter plot (right). h, A FISH image with three different probes at far-separating loci on chromosome 6 (left), the comparison of pair-wise distances of these loci in experiment and models (middle), and their relative radial positions in experiment and models (right). Numbers of regions in each boxplot for models: 10,000, for experiment: 892. In f and h, the box and the middle line in the box show the the interquartile range (IQR = Q3 – Q1) and the median. The vertical lines outside the box (and whiskers in h) extend to a maximum of 1.5*IQR beyond the box. Q1 and Q3 are the lower and upper quartile of the distribution. Outliers are shown as dots.