Abstract
Compacting chromatin within the cellular nucleus presents a significant challenge for biology. Chromosomes must be both condensed and spatially organized to enable essential processes such as transcription and replication. Chromosome conformation capture experiments (e.g., Hi-C) provide valuable information about the spatial organization and, therefore, the connectivity between different genomic regions. These experiments inspired polymer models that describe the physical mechanism of the chromosomal energy landscape. The Full-Inversion Chromatin model (FI-Chrom), a data-driven approach for modeling genome organization, uses Hi-C contact maps to infer pairwise interaction potentials between all chromosomal loci. It combines Graphics Processing Unit (GPU)-accelerated simulations with efficient training of tens of millions of parameters derived from the maximum-entropy principle to determine 3D structures of chromosomes that accurately reproduce Hi-C-like data. FI-Chrom does not make any a priori assumptions regarding chromosome architecture, making it applicable to any chromosome conformation capture experiment. Its derived structural ensembles capture all essential features from the short- and long-range interactions of typical chromosome organization, such as segregated compartments, chromosome territories, and fully or partially formed loops. Although Hi-C contains only structural information, FI-Chrom extends these data by revealing an emergent dynamical mechanism encoded in the inferred energy landscape. For example, simulations show that chromatin loops are not static architectural features but rather transient structural elements. Statistical analyses further indicate that loops confined within a single compartment occur more frequently than those spanning multiple compartments, highlighting the dynamic and compartment-dependent nature of chromatin organization.