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. Empowered by adversarial training, our method can restore fine-grained details similar to those in high-resolution Hi-C matrices, boosting accuracy in chromatin loops identification and TADs detection, and outperforms the state-of-the-art methods in accuracy of prediction. Finally, application of DeepHiC to Hi-C data on mouse embryonic development can facilitate chromatin loop detection with higher accuracy. We develop a web-based tool (DeepHiC, http://sysomics.com/deephic) that allows researchers to enhance their own Hi-C data with just a few clicks. Author summaryWe developed a novel method, DeepHiC, for enhancing Hi-C data resolution from low-coverage sequencing data using generative adversarial network. DeepHiC is capable of reproducing highresolution (10-kb) Hi-C data with high quality even using 1/100 downsampled reads. Our method outperforms the previous methods in Hi-C data resolution enhancement, boosting accuracy in chromatin loops identification and TADs detection. Application of DeepHiC on mouse embryonic development data shows that enhancements afforded by DeepHiC facilitates the chromatin loops identification of these data achieving higher accuracy. We also developed a user-friendly web server (http://sysomics.com/deephic) that allows researchers to enhance their own low-resolution Hi-C data (40kb-1Mb) with just few clicks. The high-throughput chromosome conformation capture (Hi-C) technique [1] is a genome-wide technique used to investigate three-dimensional (3D) chromatin conformation inside the nucleus. It has facilitated the identification and characterization of multiple structural elements, such as the A/B compartment [1], topological associating domains (TADs) [2, 3], enhancer-promoter loops [4] and stripes [5] over recent decades. In practice, Hi-C data is conventionally stored as a pairwise read count matrix , where is the number of observed interactions (read-pair count) between × genomic regions and , and the genome is partitioned into fixed-size bins (e.g., 25 kb). Bin size (i.e., resolution), is a crucial parameter for Hi-C data analysis, as it directly affects the results of downstream analysis, such as predictions of enhancer-promoter interactions [6-11] or identification of TAD boundaries [6, 12-16]. Depending on sequencing depths, the size of commonly used bins ranges from 1 kb to 1 Mb. Because of the high cost of sequencing, most available Hi-C datasets have relatively low resolution [17], which limits their application in studies of genomic regulatory elements. Sequencing high-res...

Spatial density of open chromatin: an effective metric for the functional characterization of topologically associated domains

“…To assess the degree that known critical factors of Hi-C data, including sequencing depth, resolution and different normalization methods may influence the SDOC value, we calculated SDOC on Hi-C data of lower coverage by downsampling the original Hi-C data of GM12878 cell line to 1/16 as described in recent studies [ 15 , 16 ]. Correlation of SDOC calculated based on original Hi-C and downsampled Hi-C is above 0.99 ( Figure S4A ), showing relatively weak influence of sequencing depth on SDOC.…”

Application of Hi-C and other omics data analysis in human cancer and cell differentiation research

“…Liu Q et al [83] proposed hicGAN to enhance low-resolution Hi-C data through Generative Adversarial Networks (GAN). Same as hicGAN [83] , in 2020, Hong, Hao et al [84] developed the DeepHiC method, which can reproduce high-resolution Hi-C data from down-sampled reads as low as 1%. Zhilan L et al [85] developed SRHiC based on the ResNet and WDSR model.…”

VEHiCLE: a Variationally Encoded Hi-C Loss Enhancement algorithm

“…Most of the previously proposed loss functions for developing Hi-C enhancement networks draw upon loss functions prolific in the fields of computer vision [5][6][7][8][9] . While there are certainly advantages to these strategies, they derive from assumed similarities between the tasks of image superresolution and Hi-C superresolution.…”

“…Previous work on Hi-C super resolution consistently used network input window sizes of 0.4Mb x 0.4Mb at 10kb resolution, requiring networks to split chromosome contact maps into 40x40bin matrices [5][6][7][8][9] . While this strategy has seen relative success, a major disadvantage is that certain important features of Hi-C such as TADs can span ranges larger than 0.4Mb, meaning that it is impossible for previous networks to explicitly encode important information about TAD organization.…”

“…The constants and are used to stabilize the metric while the means, variances and covariances are computed via a Gaussian filter. We use the implementation of SSI developed by Hong et al 8 (DeepHiC) keeping their default values for the size of sub-windows and variance value of gaussian filter at 11 and 3 respectively.…”