SHICEDO: single-cell Hi-C data enhancement with reduced over-smoothing
View abstract on PubMed
Summary
This summary is machine-generated.SHICEDO enhances sparse single-cell Hi-C data using a deep learning model. This improves 3D genome structure analysis by accurately imputing missing contacts and refining features like compartments and loops.
Area Of Science
- Genomics
- Computational Biology
- Molecular Biology
Background
- Single-cell Hi-C (scHi-C) is crucial for understanding 3D genome organization.
- scHi-C data suffer from sparsity and noise, posing computational challenges.
Purpose Of The Study
- Introduce SHICEDO, a deep learning model to enhance scHi-C contact matrices.
- Improve imputation of missing chromatin contacts in scHi-C data.
Main Methods
- Utilize a generative adversarial framework for data enhancement.
- Incorporate customized features and channel-wise attention to address scHi-C data characteristics.
- Leverage unique structural properties of scHi-C matrices.
Main Results
- SHICEDO outperforms existing state-of-the-art methods in simulations and real-data applications.
- Achieve superior quantitative and qualitative results in scHi-C data enhancement.
- Enable more precise delineation of A/B compartments, TAD-like domains, and chromatin loops.
Conclusions
- SHICEDO effectively enhances scHi-C data quality and resolution.
- Facilitates more accurate analysis of 3D genome structures.
- Publicly available tool for the research community.