1)patterns of genetic variability in human mitochondria show evidenc.pdf
Recomb2016poster
1. UNRAVELING HUMAN GENOME ARCHITECTURE WITH POPULATION-BASED MODELING
Nan Hua, Long Pei, Ke Gong, Harianto Tjong, Wenyuan Li, Chao Dai, Qingjiao Li, Xianghong Jasmine Zhou, and Frank Alber
Molecular and ComputaHonal Biology, Department of Biological Sciences, University of Southern California, 1050 Childs Way, Los Angeles, CA
ConformaHon capture technologies (e.g. Hi-C) chart physical interacHons between chromaHn regions on a genome-wide scale. However, the structural variability
of the genome between cells poses a great challenge to interpreHng ensemble-averaged Hi-C data. Here, we present a novel and improved probabilisHc approach
for deconvoluHng Hi-C data into a model populaHon of disHnct diploid 3D genome structures, which facilitates the detecHon of chromaHn interacHons likely to
co-occur in individual cells. Using human lymphoblastoid cells, we model the whole diploid genome in TAD (topological associated domain) resoluHon. The
populaHon-based models reveal disHnct posiHonal preferences for TADs with different chromaHn epigeneHc states in the nucleus. Besides centromere and
telomere regions, we observed several loci that act to bridge mulHple chromosomes and tend to reside in the nuclear interior. In addiHon we performed
clustering analysis of chromosome structures. Our populaHon-based method and analysis provides an important tool for revealing novel insights into the key
factors shaping the spaHal genome organizaHon by providing a flexible framework for data-driven genome structure modeling.
Deconvolu>on of conforma>on capture data
into a popula>on of 3D genome structures
A-step: an efficient heurisHc strategy to esHmate
contact indicaHon tensor W by using informaHon from
the structure populaHon generated in the previous M-
step.
argmax{log( , | )}←
W
W A W X
argmax{log( , | )}←
X
X A W X
M-step: simulated annealing dynamics and conjugate
gradient opHmizaHons to generate a populaHon of 3D
genome structures X.
AcHvaHon Distance CorrecHon: we use a heurisHc
method to infer the minimal possible constraints P(k)
towards opHmizing in the final structure populaHon, by
using the previous contact profiles P(k-1).
expected constrained nonconstrainedN N N= +
( )
1
ij ijk
ij
ij
A Q
P
Q
−
=
−
( 1)
( 1)
1
k
ij ij
ij k
ij
A P
Q
P
−
−
−
=
−
Abstract
Method Similarity Variability
Corrected Model Constraints
Chromosome TADs posi>on valida>on
Structure popula>on recapitulates contact
profiles in Hi-C data
Several loci tend to reside in the nuclear
interior
Learning structural heterogeneity of
chromosome by clustering
The substanHal heterogeneity of spaHal genome
organizaHon could be revealed by applying clustering
analysis on our structural populaHon. We implement
dimensionality reducHon and clustering methods to
idenHfy the underlying structural variability using
chromosome6 in GM12878 cell as an example.
✓We generated a populaHon of 3D structures at TAD
level resoluHon that correctly predicts features of the
lymphoblastoid genome.
✓✣hromosome spaHal posiHoning and fits nicely with
experiment data.
✓Our models revealed some genomic loci are relaHvely
interior and may bridge the mulHple chromosomes
interacHons
✓We showed the model structures has significant
variability within a populaHon of cells
Conclusion
Reference
s 1.Rao S, et al. (2014) A 3D Map of the Human Genome at Kilobase
ResoluHon Reveals Principles of ChromaHn Looping. Cell, 159(7),
2.Tjong H, et al. (2016) PopulaHon-based 3D genome structure analysis
reveals driving forces in spaHal genome organizaHon. Proceedings of the
Na3onal Academy of Sciences, 201512577.
3.Kind J, et al.(2015) Genome-wide Maps of Nuclear Lamina InteracHons
in Single Human Cells. Cell, 163(1), 134–147.
4.Nagano T, et al. (2013) Single-cell Hi-C reveals cell-to-cell variability in
chromosome structure. Nature 502(7469):59–64.