Fig. 1

Simulation of cancer Hi-C data. a. In Hi-C data from diploid cells, the contact frequency measured between two loci i and j is equal to the sum of 2 cis interactions (black solid lines) occurring within an individual allele and of 2 trans interactions between homologous chromosomes (transH, black dashed lines). In addition, the contact frequency observed in trans between loci i and k is the sum of 4 interactions between non homologous chromosomes (red dashed lines). b. In the context of segmental rearrangement, these properties can be extended and generalized if loci i and j belong to the same DNA segment, or to different segments (see “Methods” section and Additional file 1: Figure S1). c. Simulation of cancer Hi-C data from normal diploid (C) data by calculating the scaling factor matrix (p). Colors in scaling factor matrix represent the level of gains (red) and loss (green) to simulate. For each interaction \(C^{\text {sim}}_{ij}\), the simulated count is finally estimated using a binomial down-sampling method (see “Methods” section). d. Intra-chromosomal maps of chromosome 1 and 2 before (top) and after (bottom) simulation of copy number changes. Copy number effects are characterized by blocks of high/lower signal. Overall, the simulation conserves the structure and the counts/distance properties of the Hi-C maps. e. Validation of the simulation model using Hi-C data from MCF10A cell line from which we simulated the expected copy number of MCF7 cancer cell line (MCF7 simulated). The mean O/E (Observed/Expected) counts per block of copy number of intra (cis) and inter-chromosomal (trans) maps at 1 Mb resolution is represented. Looking at the intra-chromosomal maps of chr3 and 8 demonstrates that our model efficiently simulates large copy number events. f. 1D genome-wide profiles of near-diploid MCF10A, simulated MCF7 and real MCF7 Hi-C data. MCF7 gain and losses are represented in red and green