|
Activity Number:
|
142
- Recent Development in Computational Biology and Bioinformatics
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Statistics in Genomics and Genetics
|
|
Abstract #316895
|
|
|
Title:
|
Deep Generative Modeling of Single Cell Chromatin Conformation Capture Data
|
|
Author(s):
|
Sunduz Keles* and Siqi Shen and Ye Zheng
|
|
Companies:
|
University of Wisconsin, Madison and University of Wisconsin and Fred Hutchinson Cancer Research Center
|
|
Keywords:
|
single cell;
chromatin conformation capture;
denoising;
deep learning;
generative modeling
|
|
Abstract:
|
Transcriptional regulatory programs of mammalian cells are influenced by the three-dimensional organization of chromatin. Distal regulatory elements interact with promoters through chromatin loops and impact development and tumorigenesis. The advent of single-cell sequencing technologies accompanied by experimental advances in mapping of 3D genome organization led to large-scale generation of single cell Hi-C data to profile higher order chromosomal structures among individual cells. Data from single cell technologies are noisy, and extremely sparse and have low counts leading to analyses based on these raw data to underestimate the biological signal within and across cells. In addition, scHi-C data inherits systematic biases such as genomic distance bias and the neighborhood effect specific to bulk Hi-C. We address these challenges by a versatile deep generative model with key components capturing the unique aspects of the scHi-C data and an established noise model for Hi-C data. Multi-model benchmarking of our model indicates that in addition to unbiased denoising and dimension reduction, this model enables robust identification of marker long-range interactions.
|
Authors who are presenting talks have a * after their name.
