Physical modeling of the genome: structural insights and biological consequences

Abstract

The genome is organized within a nucleus where chromosomes fold into an ensemble of conformations. This spatial arrangement within the nucleus is critical for regulating gene expression and other DNA-templated processes. Chromosome conformation capture techniques such as Hi-C provide information about the genome architecture by creating 2D contact maps. Recently, measuring efforts were expanded to several organisms, cell lines, tissues, and cell cycle phases, in which obtaining high-quality maps is still challenging. These contact maps serve as an essential input for top-down theoretical models. Aided by the under-development chromatin folding and structure theory, we create a framework using polymer theory enriched with the Maximum Entropy Approach to learn and understand the chromosome spatial organization. To enhance the training of the energy functions, we included a combined machine learning minimization method that allows us to speed up the modeling, even for large systems. We use our developed platform, Open-MiChroM, to perform fast simulations and training. The models generated are precise compared to the experimental Hi-Cs, and the 3D structures ensemble is consistent with the crystal liquid theory for chromosomes. Additionally, we are able to predict important features of chromosome organization, such as the phase separation between chromatin types and the formation of chromosome territories. This novel modeling allows the exploration of a broad spectrum of 3D genome organizations on different organisms, cell lines, and cell phases.