Enhancing Hi-C resolution data with computational construction of three-dimensional chromatin ensemb
Angel, J.C., B.Sc. (Columbia University, New York Genome Center) El Amraoui, N., M.Sc. (New York Genome Center) Gürsoy, G., Ph.D. (Columbia University, New York Genome Center)
While Hi-C has emerged as a pivotal technique to study the three-dimensional genome organization, its inherent shortcomings, such as high sequencing cost, compromise its resolution. This restricts a comprehensive analysis, resulting in the loss of potential biological insights such as detection of gene-enhancer interactions. Here we introduce a novel and efficient algorithm that utilizes importance sampling to transform low-resolution interaction frequencies into high-resolution ensembles of three-dimensional chromatin chains. This is achieved by generating chromatin chains one monomer at a time by ensuring that the imposed physical constraints from chromatin interactions are fully satisfied. We demonstrate the efficacy of our method in reconstructing high-resolution Hi-C data from as few as 1% of the original Hi-C matrix interactions. Moreover, our algorithm demonstrates a superior performance over existing deep learning methods, achieving over 95% of accuracy in reconstructing high-resolution chromatin maps, thus providing a cost-effective and efficient solution to study the intricacies of genome organization and the processes it influences.
2023 MidAtlantic Bioinformatics Conference Lightning Talk