Diffusion-based generation of gene regulatory networks from scRNA-seq data with DigNet

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Gene regulatory networks (GRNs) are crucial for cellular specificity, but their reconstruction from gene expression data is challenging.
• Existing methods struggle with cell-specific GRN reconstruction in precise cellular and genetic contexts.
• High-throughput single-cell RNA sequencing (scRNA-seq) data offers unprecedented resolution but presents challenges due to noise and sparsity.

Purpose of the Study:

• To develop a novel computational model for generating cell-specific GRNs from scRNA-seq data.
• To address the limitations of existing methods in reconstructing complex and context-specific gene regulatory architectures.
• To provide a robust and accurate tool for inferring gene regulatory relationships at the single-cell level.

Main Methods:

• Proposed DigNet, a discrete diffusion generation model embedding network generation into a multistep recovery procedure with Markov properties.
• Utilized iMetacell integration and non-Euclidean discrete space modeling for robustness against data noise and sparsity.
• Employed a unique multistep diffusion procedure to ensure compatibility between global network structures and regulatory modules.

Main Results:

• DigNet demonstrated superior performance compared to over a dozen state-of-the-art network inference methods in benchmark evaluations.
• The model successfully reconstructed cell-specific GRNs from scRNA-seq data, handling noise and sparsity effectively.
• DigNet provided novel insights into the immune response in breast cancer by identifying differential gene regulation in T cells.

Conclusions:

• DigNet is a powerful and effective open-source tool for generating cell-specific GRNs from scRNA-seq data.
• The model's approach ensures compatibility between global network structures and local regulatory modules.
• DigNet advances the field of GRN inference and offers valuable applications in understanding cellular specificity and disease biology.