scReGAT: Leveraging Knowledge of Regulatory Interactions to Predict Long-range Gene Regulation at Single-cell...

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Gene regulation is key to understanding cellular processes, development, and disease.
• Single-cell resolution is essential for dissecting cellular heterogeneity.
• Existing methods struggle to capture dynamic, long-range regulatory interactions.

Purpose of the Study:

• To develop a deep learning framework for reconstructing cell-specific gene regulatory networks.
• To integrate prior knowledge of regulatory elements and transcription factors with single-cell data.
• To identify dynamic regulatory rewiring and disease-associated mechanisms at single-cell resolution.

Main Methods:

• Introduced single-cell regulatory graph attention network (scReGAT), a deep learning framework.
• Constructed a knowledge-guided regulatory graph (kRG) using validated interactions and chromatin accessibility.
• Trained a Graph Attention Network (GAT) to predict gene expression and quantify regulatory contributions.

Main Results:

• scReGAT successfully recapitulated known cell-type-specific cis-regulatory element (cRE)-gene interactions across five datasets.
• Identified dynamic regulatory rewiring predicting transcriptional transitions in neuroblastoma and osteogenic differentiation.
• Uncovered candidate regulatory mechanisms for Alzheimer's, multiple sclerosis, and schizophrenia by integrating GWAS loci.

Conclusions:

• scReGAT provides a robust and generalizable framework for decoding long-range gene regulation at single-cell resolution.
• The tool can identify cell types associated with complex diseases and elucidate underlying regulatory mechanisms.
• Enables deeper insights into gene regulation crucial for development, disease, and cellular identity.