Fig. 5

a Overlap between clusters from four cell lines inferred by Arboretum-Hi-C. Each red-blue matrix shows the extent of similarity between clusters from pairs of cell lines or species as measured by − log10P value of a hypergeometric test. Diagonal elements represent clusters of the same ID and are shown in red. Off-diagonal elements are shown in blue. The intensity of red and blue is proportional to the extent of similarity between pairs of clusters. b The distribution of conservation score between pairs of cell lines estimated from multiple random initializations of Arboretum-Hi-C and independent spectral clustering. The conservation score for Arboretum-Hi-C was defined as the sum of diagonal elements minus off-diagonal elements of the matrices from (a). Because independent spectral clustering does not give a mapping of cluster assignments across data sets, we first matched cluster IDs based on maximal overlap of regions using the Hungarian algorithm (Methods). hESC human embryonic stem cell, hIMR90 IMR90 human fibroblast, mCortex mouse cortex, mESC mouse embryonic stem cell