2020
DOI: 10.1371/journal.pcbi.1007287
|View full text |Cite
|
Sign up to set email alerts
|

DeepHiC: A generative adversarial network for enhancing Hi-C data resolution

Abstract: Hi-C is commonly used to study three-dimensional genome organization. However, due to the high sequencing cost and technical constraints, the resolution of most Hi-C datasets is coarse, resulting in a loss of information and biological interpretability. Here we develop DeepHiC, a generative adversarial network, to predict high-resolution Hi-C contact maps from low-coverage sequencing data. We demonstrated that DeepHiC is capable of reproducing high-resolution Hi-C data from as few as 1% downsampled reads. Empo… Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
105
0

Year Published

2020
2020
2024
2024

Publication Types

Select...
5
3
2

Relationship

0
10

Authors

Journals

citations
Cited by 74 publications
(105 citation statements)
references
References 43 publications
0
105
0
Order By: Relevance
“…The work associated with the higher level of predictions and knowledge discovery with the analytics Hao Hong et al [14] created DeepHiC whereby the high performance generative adversarial network is used. The study indicated that DeepHiC is able to mate high-resolution Hi-C data from as few down sampled reads as 1 percent.…”
Section: Review Of Existing Workmentioning
confidence: 99%
“…The work associated with the higher level of predictions and knowledge discovery with the analytics Hao Hong et al [14] created DeepHiC whereby the high performance generative adversarial network is used. The study indicated that DeepHiC is able to mate high-resolution Hi-C data from as few down sampled reads as 1 percent.…”
Section: Review Of Existing Workmentioning
confidence: 99%
“…Hi-C is commonly used to study three-dimensional genome organization. Hong et al (105) developed a GAN, namely DeepHiC, to predict high-resolution Hi-C contact maps from low-coverage sequencing data. DeepHiC can reproduce highresolution Hi-C data from as few as 1% down sampled reads.…”
Section: Application Of Gan In Bioinformaticsmentioning
confidence: 99%
“…Using Micro-C contact maps from hESC, mESC, and HFF as the prediction target, the model was trained with backpropagation [10], in which the aforementioned convolutional features were learned adaptively. Other than leveraging a number of epigenomic features, our model architecture differs from HiCPlus [11] and DeepHiC [12] which treats Hi-C contact maps as images and performs grid-convolution to improve the resolution. With the graph convolutional networks and additional epigenomic features, CAESAR not only enhances the resolution of contact maps, but also predicts the structures which are not captured by Hi-C, including polycomb repressive regions, short-range loops and stripes ( Figure 1b).…”
Section: A Deep Learning Model Imputing High-resolution Chromatin Conmentioning
confidence: 99%