|
Abstract:
|
Recent Hi-C experiments have enabled researchers to study the chromatin structures at fine resolutions as low as 1 kilobase. Mathematical algorithms aimed at reconstructing the chromatin structure are based on optimizing the structure such that pairwise distances between loci are inversely proportional to the corresponding interaction frequency. However, such methods can be sensitive to resolutions. When resolution is high, the high proportion of observed zero interaction contacts can lead to erroneous structure. To address the high sparsity of contact frequency maps, we propose an optimization-based method for reconstruction of spatial organization of the chromatin penalizing on distances between loci from different topological associated domains (TAD). These domains are observed to be associated with CTCF-binding factors and are independent of the resolution. Scalability of the model is observed at different resolutions in simulation studies, where the structure of the TADs is preserved. The method is applied to the Rao et al., 2014 data at various resolutions (1 kb - 100 kb) to study reproducibility of the spatial map.
|
