Spatial Genome Architecture in Development & Disease

Algorithms for inferring haplotype-specific chromatin contact maps in cancer using Genome Architecture Mapping


DNA and proteins together form chromatin, which is packaged in the nucleus in a highly structured way. This structure plays an important role in tissue development and in maintaining tissue integrity and is frequently disrupted in cancer and other diseases. Despite its importance, identifying changes in chromatin structure in cancer is challenging. Cancer genomes are highly unstable and evolve in the course of disease progression, rearranging their chromosomes and creating changes in chromosome copy-number. These copy-number changes confound the analysis of chromatin structure and to understand their interplay both high-resolution copy-number data and high-resolution chromatin conformation data are required. In addition, since healthy cells contain two parental copies of each chromosome (two haplotypes), it would be highly beneficial to be able to distinguish these haplotypes when analysing copy-number and chromatin structure.Genome Architecture Mapping (GAM) is a new experimental method for determining chromatin structure based on cryosectioning and sequencing of cell nuclei. In our previous work we demonstrated that during the sectioning process, GAM locally mostly captures one of the two haplotypes. This allowed us to devise an algorithm to assign genetic variants to the haplotype they originate from and to generate haplotype-specific chromatin contact maps, albeit only in genomes with a sufficiently large number of variants, such as mouse. Independently, on cancer sequencing data, we have developed phasing algorithms that allowed us to determine haplotype-specific copy-number profiles with very high accuracy.To be able to jointly infer chromatin structure and copy-number variants in cancer in a haplotype-specific manner we propose several algorithmic advances. We first propose to extend our algorithm GAMIBHEAR to be amenable to human sequencing data with lower variant densities. Using a “co-phasing” strategy, we will be able to assign sequencing reads that do not overlap heterozygous variant positions to their parental haplotype-of-origin. This will enable us to derive high-resolution haplotype-specific chromatin contact maps in human. In the next step, we will leverage our experience in developing phasing algorithms for copy-number variants to develop a novel algorithm for detecting haplotype-specific copy-number changes in GAM data. Finally, this algorithm will be extended to also include genomic rearrangements which are copy-number neutral, such as balanced translocations and inversions. This set of algorithms and tools will ultimately enable the routine application of GAM to clinical cancer specimen.

  • Dr. Roland Schwarz,


    Universitätsklinikum Köln
    Centrum für Integrierte Onkologie (CIO)
    Kerpener Straße 62
    50937 Köln