Sparse 3D reconstructions.

<p>The figure contains boxplots of the benchmark test results for each of the 5 model categories (H1-4, R1). 20 models were generated from each type. Random models were generated by shuffling the coordinates of the original <i>S</i>. <i>cerevisiae</i> Hi-C distance map. Results that are distributed significantly above or below the random reference models (their median marked with a line) according to Wilcoxon rank-sum (one-tail), are denoted with a star or more with respect to their significance level (one star for p<0.05, two for p<0.01, three for p<0.001). We observe that on most tests all Hi-C model types obtain similar results, with some tests showing a gradual increase in signal gain with the number of Hi-C interactions. <b>(A)</b> Optimization objective function of the reconstructed solution, normalized with respect to random models with similar properties. <b>(B)</b> Average Spearman’s correlation between the pairwise distances in each model (9.1x10⁵ points) with the other reconstructions generated in its category. <b>(C)</b> Centromere co-localization, measured in normalized set distance (NSD), expected to be lower/greater than 1 for co-localized/dispersed sets, respectively. <b>(D)</b> Telomere radius from the center of the nucleus. <b>(E)</b> Ratio of the average <i>cis</i> (intra-chromosomal) distances between chromosome arms and <i>trans</i> (inter-chromosomal) distances. <b>(F)-(L)</b> Co-localization results for various sets of functional loci. Where the set comprises of several co-localized subsets (such as each GO term, tRNA clusters 1 and 2, etc.), the result presented is the mean of the sets’ mean distance. <b>(M)</b> Spearman's correlation between pairwise distances of genes and their coefficient of correlation of expression (<i>n</i> = 2,000 bins). <b>(N)</b> Spearman's correlation between pairwise distances of genes and their distances on a protein-protein interaction (PPI) graph (<i>n</i> = 2,000 bins). <b>(O)</b> Spearman's correlation between pairwise distances of genes and their distances according to the codon usage frequency similarity (CUFS) (<i>n</i> = 2,000 bins).