Multidimensional scaling for three-dimensional modeling of chromosome conformation

Restricted (Penn State Only)
Author:
Rieber, Lila
Graduate Program:
Bioinformatics and Genomics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
December 16, 2019
Committee Members:
  • Shaun Mahony, Dissertation Advisor/Co-Advisor
  • Shaun Mahony, Committee Chair/Co-Chair
  • Feng Yue, Committee Member
  • Ross Cameron Hardison, Committee Member
  • Qunhua Li, Outside Member
  • George H Perry, Program Head/Chair
  • Benjamin F. Pugh, Committee Member
Keywords:
  • chromosome conformation
  • gene regulation
Abstract:
Chromatin has a non-random conformation within the nucleus, which is correlated with gene regulation and expression. The 3D organization of chromatin can be studied using the sequencing-based Hi-C assay, but raw Hi-C data requires computational analysis to be interpretable. We have developed two tools based on multidimensional scaling (MDS) that can be applied to Hi-C data. miniMDS uses the hierarchical 3D organization of chromosomes to implement an efficient parallelized approximation to MDS. It is fast and uses limited memory, even at high resolution. We used miniMDS to perform whole-genome structural inference on GM12878 Hi-C data at 10-kb resolution. We also developed MultiMDS, which builds on miniMDS, to jointly infer and align structures from two Hi-C data sets. We applied MultiMDS to various Hi-C data sets: yeast grown with glucose and yeast grown with galactose, human cell types from the ENCODE project, mouse cell types, lymphoblastoid cell lines from the 1000 Genomes Project, architectural protein depletions, and in vitro neural differentiation. We quantified the contribution of compartment differences to the overall change in 3D genomic structure for each of the mammalian comparisons. We also identified compartment-independent relocalizations that were correlated with biological function, including the relocalization of galactose genes in yeast and relocalizations between mammalian cell types that were correlated with cell-type-specific histone modifications.