Abstract
The 3D organization of the genome is central to gene regulation, and phase separation has emerged as an important physical principle for this architecture. This review synthesizes how phase separation contributes to genome folding across scales, from compartmental segregation and topologically associating domains to transcriptional condensates and nucleosome arrays, with a special focus on computational advances. We organize the field into two complementary modeling paradigms: (1) physics-based simulations, spanning all-atom to coarse-grained polymer representations that reveal the mechanisms driving chromatin condensation; and (2) data-driven approaches, including machine learning, that learn structural features and regulatory interactions from high-throughput genomic and imaging data. We highlight how integrating these models with experiments clarifies the interplay among phase separation, loop extrusion, epigenetic modifications, and the intrinsic polymer properties of chromatin in genome folding. By linking microscopic molecular interactions to mesoscale and nuclear organization, these combined approaches provide mechanistic insight into normal regulation and its dysregulation in disease, and they chart a path toward predictive, non-equilibrium models of the 4D nucleome.