Free energy based high-resolution modeling of CTCF-mediated chromatin loops for human genome

Abstract

AbstractIn recent years, chromatin has been found to have considerable structural organization in the human genome with diverse parts of the chromatin interacting with each other to form what have been termed topologically associated domains (TADs). Chromatin interaction analysis by paired-end tag sequencing (ChIA-PET) is a recent protein-specific method that measures these chromatin interactions via specific interactions such as CTCF-cohesin binding proteins or RNA polymerase II interactions. Unlike high-throughput chromosome conformation capture (Hi-C), which measures unspecific binding (all against all), ChIA-PET measures specific protein-protein contact interactions; hence physical bonds that reflect binding free energies. In this work, a thermodynamic method for computing the stability and dynamics of chromatin loops is proposed. The CTCF-mediated interactions, as observed in ChIA-PET experiments for human B-lymphoblastoid cells, are evaluated in terms of a chain folding polymer model and the experimentally observed frequency of contacts within the chromatin regions. To estimate the optimal free energy and a Boltzmann distribution of suboptimal structures, the approach uses dynamic programming with methods to handle degeneracy and heuristics to compute parallel and antiparallel chain stems and pseudoknots. Moreover, multiple loops mediated by CTCF protein binding that connects together more than one chain into multimeric islands are simulated using the model. Based on the thermodynamic properties of those topological three-dimensional structures, we predict the correlation between the relative activity of chromatin loop and the Boltzmann probability, or the minimum free energy, depending also on its genomic length. The results show that segments of chromatin where the structures show a more stable minimum free energy (for a given genomic distance) tend to be inactive, whereas structures that have lower stability in the minimum free energy (with the same genomic distance) tend to be active.