Cross-cell-line conservation-resolved interactions reveal conservation-dependent relationships among CTCF loops.

The three-dimensional organization of the genome is shaped by CTCF-mediated chromatin loops, which vary widely in strength, conservation, and cell-type specificity. A longstanding model proposes that highly conserved loops form autonomously and establish a structural scaffold that supports the subsequent formation of less conserved, cell-type-specific interactions. However, genome-wide evidence for an ordered dependency among loops of different conservation levels has remained limited.Here, we tested this hierarchical model using high-resolution CTCF ChIA-PET data from eight human cell lines. We developed a two-stage predictive framework in which high-confidence loops were first identified using sequence features, chromatin context, and cross-cell-line conservation, and then augmented with neighboring-interaction features that quantify the influence of pre-existing loops at nearby CTCF anchors. Incorporation of neighboring-interaction information resulted in only modest overall improvements in predictive performance.To directly assess hierarchical dependencies, we stratified loops into eight conservation classes based on their recurrence across cell lines and systematically evaluated how neighboring interactions from each class contributed to loop prediction in others. This conservation-resolved analysis revealed a structured pattern in predictive relationships: loops within a given conservation class were most strongly predicted by neighboring loops from adjacent conservation classes. In contrast, the most highly conserved loops showed little improvement from local loop-context information.Together, these results demonstrate that the predictive value of neighboring chromatin interactions depends strongly on loop conservation and that chromatin interaction neighborhoods contain structured conservation-dependent information. More broadly, our neighboring-interaction framework provides an interpretable approach for identifying structured associations within chromatin interaction networks and for generating hypotheses regarding the organization of three-dimensional genome architecture.