HiCONA: a graph-based framework for identifying the hierarchical 3D backbone of chromatin

Understanding chromatin architecture requires robust methods that capture its nested hierarchy across scales. Sequencing-based approaches such as Hi-C provide genome-wide contact maps, but their inherent sparsity and noise limit structural interpretation. Moreover, existing computational tools often focus on either local features or global patterns, hindering an integrated view. Here, we present HiCONA (Hi-C Organization with Network Analysis), a generative Bayesian framework that infers multiscale chromatin organization using the Nested Stochastic Block Model. HiCONA models chromatin as a hierarchical network and assigns contact importance scores, enabling denoising, feature discovery, and functional interpretation. Applied to both sequencing and imaging datasets, HiCONA accurately identifies known structures across scales, recovers important contact patterns including microcompartments, and recapitulates architectural changes after cohesin-depletion. HiCONA is implemented as an accessible Python tool, HiCONA facilitates interpretable, unified analysis of chromatin architecture and opens the door to integrating Hi-C data with other omics modalities.