hicGAN infers super resolution Hi-C data with generative adversarial networks

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Hi-C technology provides genome-wide insights into 3D chromatin conformation by measuring DNA contact frequencies.
• High-resolution Hi-C data are essential for accurate identification of genomic structures like topologically associating domains (TADs) and chromatin loops.
• The high cost of sequencing limits the availability of high-resolution Hi-C data across diverse cell types, necessitating computational solutions.

Purpose of the Study:

• To introduce hicGAN, a novel open-sourced framework for inferring high-resolution Hi-C data from low-resolution data.
• To apply generative adversarial networks (GANs) for the first time in 3D genome analysis to enhance Hi-C data resolution.
• To provide a computational approach for generating high-quality Hi-C data where experimental high-resolution data is unavailable.

Main Methods:

• Utilized generative adversarial networks (GANs) within the hicGAN framework.
• Trained the model to infer high-resolution Hi-C contact matrices from low-resolution input data.
• Validated the performance by comparing generated high-resolution matrices against original high-resolution Hi-C data.

Main Results:

• hicGAN effectively enhances the resolution of low-resolution Hi-C data.
• Generated Hi-C matrices demonstrate high consistency with experimentally derived high-resolution Hi-C matrices.
• The framework successfully infers detailed 3D genome conformation, enabling better analysis of gene regulation.

Conclusions:

• hicGAN offers a powerful computational method to overcome the limitations of sequencing costs for obtaining high-resolution Hi-C data.
• This approach facilitates the study of 3D genome organization and its functional implications in various biological contexts.
• The open-sourced nature of hicGAN promotes wider accessibility and application in 3D genomics research.