Activity Number:
|
428
- Clustering and Dimension-Reduction Methods: From Omics to Single-Cell Data
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 10, 2022 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
Abstract #323444
|
|
Title:
|
Denoising and Inference for Single Cell Chromosome Conformation Capture Data (ScHi-C) by Large-Scale Unbiased Tensor Decomposition
|
Author(s):
|
Kwangmoon Park* and Sunduz Keles
|
Companies:
|
University of Wisconsin-Madison and University of Wisconsin, Madison
|
Keywords:
|
Statistical Genomics;
Single cell sequencing;
scHi-C;
Tensor model;
High-dimensional Statistics
|
Abstract:
|
The three-dimensional (3D) configuration of the genome has critical implications for gene regulation by bringing distal regulatory elements in close proximity of the genes. Combination of advancements in sequencing of single cells and whole genome profiling of long-range interactions enabled single-cell resolution investigations of 3D chromatin interactions by scHi-C data. Although scHi-C data naturally lends itself to a high dimensional tensor structure, initial approaches thus far have not leveraged this property, potentially owing to the extreme sparsity and inherent technology-dependent biases of the data structure. We develop a Debiased Joint Block Term Decomposition (DJBTD) that accommodates the well-known genomic distance bias of chromatin conformation capture technologies and is able to simultaneously handle data from all the chromosomes. Our data-driven simulations and large-scale data analysis show that DJBTD provides accurate cell-type specific mean contact map estimates and cell clustering. We further utilize these contact map estimates for inferring cell-type specific contacts and topologically associating domains.
|
Authors who are presenting talks have a * after their name.