Joint inference and alignment of genome structures enables characterization of compartment-independent reorganization across cell types

Abstract

Abstract Background Comparisons of Hi–C data sets between cell types and conditions have revealed differences in topologically associated domains (TADs) and A/B compartmentalization, which are correlated with differences in gene regulation. However, previous comparisons have focused on known forms of 3D organization while potentially neglecting other functionally relevant differences. We aimed to create a method to quantify all locus-specific differences between two Hi–C data sets. Results We developed MultiMDS to jointly infer and align 3D chromosomal structures from two Hi–C data sets, thereby enabling a new way to comprehensively quantify relocalization of genomic loci between cell types. We demonstrate this approach by comparing Hi–C data across a variety of cell types. We consistently find relocalization of loci with minimal difference in A/B compartment score. For example, we identify compartment-independent relocalizations between GM12878 and K562 cells that involve loci displaying enhancer-associated histone marks in one cell type and polycomb-associated histone marks in the other. Conclusions MultiMDS is the first tool to identify all loci that relocalize between two Hi–C data sets. Our method can identify 3D localization differences that are correlated with cell-type-specific regulatory activities and which cannot be identified using other methods.

Keywords

• Hi–C
• Structural inference
• Gene regulation