ChromatinHD connects single-cell DNA accessibility and conformation to gene expression through scale-adaptive machine...

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• Gene regulation is a multiscale process, but current methods struggle to fully capture this complexity in single-nucleus accessibility data.
• Existing approaches often rely on peak-calling or window-based analyses, potentially missing functional regulatory information.
• There is a need for scale-adaptive machine learning models to analyze raw chromatin accessibility data effectively.

Purpose of the Study:

• To develop ChromatinHD, a novel pair of scale-adaptive models for analyzing single-nucleus accessibility data.
• To link chromatin accessibility regions directly to gene expression without preprocessing steps like peak-calling.
• To identify differentially accessible chromatin and understand its relationship with gene regulation across various genomic scales.

Main Methods:

• Developed ChromatinHD, a pair of scale-adaptive machine learning models utilizing raw ATAC-seq accessibility data.
• The models directly link chromatin accessibility patterns to gene expression levels.
• Incorporated analysis of ATAC-seq fragment length variations to detect transcription factor binding density.

Main Results:

• ChromatinHD consistently outperformed existing peak and window-based methods in linking accessibility to gene expression.
• The models captured numerous functional accessibility changes, both within and outside of predicted cis-regulatory regions.
• ChromatinHD identified collaborating regulatory regions and their genomic conformations influencing gene expression, and detected dense transcription factor binding via fragment length analysis.

Conclusions:

• ChromatinHD provides a data-driven approach to understand chromatin accessibility at multiple scales and its impact on gene expression.
• The scale-adaptive models offer superior performance by analyzing raw data and capturing subtle regulatory features.
• This computational suite advances the analysis of epigenomic data for a deeper understanding of gene regulation.