Connecting high-resolution 3D chromatin organization with epigenomics.

Area of Science:

• Genomics
• Computational Biology
• Epigenetics

Background:

• Chromatin conformation capture technologies are advancing, enabling higher resolution 3D chromatin organization studies.
• Understanding the relationship between fine-scale 3D chromatin structure and epigenomic states is crucial for human cell biology.

Purpose of the Study:

• To develop a deep learning model (CAESAR) that maps epigenomic features to high-resolution 3D chromatin organization.
• To accurately predict fine-scale chromatin structures, including loops and stripes, beyond the capabilities of traditional Hi-C methods.
• To impute high-resolution 3D chromatin contact maps for various human tissues and cell lines.

Main Methods:

• Utilized publicly available Micro-C datasets.
• Developed a deep learning model named CAESAR.
• Integrated epigenomic datasets from ENCODE and Roadmap Epigenomics Project.

Main Results:

• CAESAR accurately predicts fine-scale chromatin structures like loops and stripes.
• Successfully imputed high-resolution 3D chromatin contact maps for 91 human tissues and cell lines.
• Identified spatial interactions between genes and regulatory elements in imputed maps.

Conclusions:

• CAESAR demonstrates potential in coupling transcriptional regulation with high-resolution 3D chromatin organization.
• The model enables the imputation of detailed 3D chromatin contact maps, facilitating further biological discoveries.
• This work advances the study of genome organization and its functional implications in human cells.