Related Experiment Video
Updated: Aug 6, 2026
07:12
Whole-cell Super-Resolution Imaging via DNA-PAINT on a Spinning Disk Confocal with Optical Photon Reassignment
Published on: January 6, 2026
HiC2Self: Self-supervised denoising for bulk and single-cell Hi-C contact maps
Rui Yang1,2,3, Alireza Karbalayghareh1,4, Christina S Leslie1,2
1Computational and Systems Biology Program, Memorial Sloan Kettering Cancer Center, New York, NY, USA.
Science Advances
|July 23, 2026
Summary
HiC2Self is a new self-supervised framework that effectively denoises low-coverage Hi-C contact maps. This tool reconstructs 3D genome structures like TADs and loops from various data types, improving single-cell analyses.
Area of Science:
- Genomics
- Molecular Biology
- Bioinformatics
Background:
- Hi-C assays reveal 3D genome organization but single-cell Hi-C faces challenges with low data coverage and sparsity.
- Existing methods struggle with generalization and data sparsity in 3D genome studies.
Purpose of the Study:
- To introduce HiC2Self, a self-supervised framework for denoising Hi-C contact maps using low-coverage data.
- To enable accurate reconstruction of 3D chromatin structures from diverse genomic datasets.
Main Methods:
- Developed HiC2Self, a self-supervised learning framework for processing Hi-C data.
- Applied HiC2Self to bulk Hi-C, Micro-C, and single-nucleus methyl-3C data.
- Validated HiC2Self's ability to reconstruct topologically associating domains (TADs) and loops.
Main Results:
- HiC2Self successfully reconstructs TADs and loops from low-coverage bulk Hi-C data without supervised learning limitations.
- Accurate reconstruction of significant loops from 1-kilobase resolution Micro-C data was achieved.
- Local TAD structures were reconstructed from single-nucleus methyl-3C data with as few as 50 cells.
- Enabled examination of single-cell 3D structures at 50-kilobase resolution.
Conclusions:
- HiC2Self provides a robust method for denoising various 3D contact map data, including bulk, pseudobulk, and single-cell Hi-C.
- The framework enhances the analysis of 3D genome organization, particularly in low-coverage and single-cell scenarios.
- HiC2Self is a versatile tool for advancing 3D genomics research.
