Abstract:
|
This paper develops a statistical framework for comparing populations of 3D chromosomal structures. The two populations can result from, for example, estimation of chromosomal structures from wild type (WT) and gene knockout (KO) contact matrices. The actual estimation is performed using a package called SIMBA3D. This algorithm takes a contact matrix as input and generates random estimates of chromosomal structure due to random initialization. Treating them as samples from a population, we compare underlying populations using elastic shape metric between chromosome structures in a pairwise fashion. Two types of comparisons are performed: (i) A global comparison where full shapes of chromosomes are considered at once, and we discover significant differences between WT and KO populations. (ii) A local comparison where chromosomes are segmented into smaller structures and one compares these segments across populations. Once again we discover some local regions of large differences across populations.
|