Figure 1.

(A) Pipeline of the IC-Finder algorithm: a Hi-C map (left bottom) is transformed into a distance map (left top). Construction of the hierarchical organization (right) is performed using agglomerative hierarchical clustering on the distance map. At a given stage of the clustering, a choice to stop merging consecutive clusters is made regarding the heterogeneity σ_norm inside the candidate merged cluster. (B) Prediction robustness estimated using Poisson resampling of the original map (left): probability p_d that two loci are predicted to ‘colocalize’, i.e. to belong to the same interaction domain (right top) and the probability p_b that a locus is predicted as a domain boundary (right bottom). (C) Different types of outputs given by IC-Finder: TAD boundaries determined by the program in default mode, and p_d, p_b and TAD boundary for different hierarchy of folding when the corresponding options are activated. Examples are given for the region 12–14.5 Mb of fly chromosome 3R (10).