Abstract
Genome-wide chromosomal contact maps are widely used to uncover the 3D organisation of genomes. They rely on the collection of millions of contacting pairs of genomic loci. Contact frequencies at short range are usually well measured in experiments, while there is a lot of missing information about long-range contacts. We propose to use the sparse information contained in raw contact maps to determine high-confidence contact frequency between all pairs of loci. Our algorithmic procedure, Boost-HiC, enables the detection of Hi-C patterns such as chromosomal compartments at a resolution that would be otherwise only attainable by sequencing a hundred times deeper the experimental Hi-C library.