Inference of gene coexpression networks from single-cell transcriptome data based on variance decomposition analysis.

Area of Science:

• Genomics
• Computational Biology
• Developmental Biology

Background:

• Gene regulation differences across cell types and developmental stages dictate cellular functions and disease pathology.
• Single-cell transcriptome data presents challenges for gene coexpression network reconstruction due to noise and sparsity.
• Understanding cell type-specific gene coexpression is vital for biological and medical research.

Purpose of the Study:

• To present an efficient method, gene coexpression networks via variance decomposition analysis (GCNVDA), for inferring gene regulatory mechanisms from single-cell transcriptome data.
• To address the challenges of noise and data sparsity in single-cell data analysis for network inference.

Main Methods:

• Developed GCNVDA, a novel method incorporating multiple variability sources, including gene-level variance ($G$) and residual errors ($E$).
• Applied GCNVDA to three real-world single-cell datasets for network inference and analysis.

Main Results:

• GCNVDA demonstrated superior sensitivity and specificity compared to existing methods in identifying tissue- or state-specific gene regulations.
• The method successfully identified functional modules crucial for biological processes, such as embryonic development.

Conclusions:

• GCNVDA provides an effective approach for reconstructing gene coexpression networks from single-cell data.
• Findings offer new insights into cell-specific regulatory mechanisms, advancing developmental biology and disease pathology research.