hicream: a flexible framework to identify significantly different regions in Hi-C data.

Abstract

The three-dimensional (3D) organization of the genome plays a central role in many biological processes, and its disruption can critically impair cellular function. Detecting such changes is therefore crucial and can be achieved through differential analysis of Hi-C data. By comparing Hi-C data across biological conditions, differential Hi-C analysis aims to identify statistically significant and biologically relevant changes in chromatin organization. However, most existing methods either target a single predefined type of structure (e.g., TADs, loops, or compartments), thereby limiting their ability to uncover novel or complex structural variations, or produce highly local and disconnected results that are difficult to interpret in terms of higher-order genome organization. We introduce hicream, a novel framework for differential Hi-C analysis that identifies regions of arbitrary shape, allowing to uncover new differential structures. By combining pixel-level differential analysis and data-driven clustering to define candidate regions of the Hi-C interaction matrix, our approach provides an interpretable measure of the differential signal for each region through a post hoc inference strategy. The resulting differential regions can be explored using an interactive visualization interface. hicream is available at https://cran.r-project.org/package=hicream.