ChromInSight: Revealing DNA Double‐Strand Breaks Through Chromatin Structural Insights With an Interpretable Graph Neural Network Framework

Abstract DNA double‐strand breaks (DSBs) represent one of the most severe forms of genomic damage. Although substantial progress has been made in elucidating general patterns associated with DSBs, the influence of 3D chromatin structure on DSB formation remains underexplored, particularly concerning its spatial configuration. Here, the ChromInSight framework is introduced. Using standardized datasets,Hi‐DSB is developed and deployed in ChromInSight, a genome‐wide DSB prediction model based on graph contrastive learning (GCL), and applied advanced interpretability techniques to identify DSB‐associated genomic patterns. The findings reveal that the spatial cluster‐scene between hub nodes and DSB sites is predominantly shaped by the 3D conformation of chromatin, rather than by linear genomic distance. This phenomenon is validated at both the Loop and topologically associating domain (TAD) levels and proposed a “spatial isolation – damage containment” hypothesis, which illustrates the genome strategy for managing damage. These findings support the role of 3D genome architecture in genomic instability. Consequently, the framework provides a powerful tool for investigating the intricate relationship between chromatin structure and genomic stability.