Joint gene network construction by single-cell RNA sequencing data.

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Gene regulatory network (GRN) analysis offers deeper insights into genetic architectures of human diseases than single-gene approaches.
• Single-cell RNA sequencing (scRNA-seq) enables finer resolution GRN construction but presents challenges due to data sparsity.
• Existing methods often fail to construct GRNs across multiple conditions simultaneously.

Purpose of the Study:

• To develop a novel method for constructing joint gene regulatory networks from scRNA-seq data.
• To address the sparsity and zero-inflation inherent in scRNA-seq data for robust network inference.
• To enable the analysis of GRNs across different but related biological conditions at single-cell resolution.

Main Methods:

• Proposes Joint Gene Networks with scRNA-seq data (JGNsc) within the Gaussian graphical models (GGMs) framework.
• Introduces a hybrid imputation strategy combining a Bayesian zero-inflated Poisson model and iterative low-rank matrix completion.
• Applies a nonparanormal transformation to imputed data before constructing joint GGMs.

Main Results:

• JGNsc effectively imputes zero-inflated scRNA-seq data, overcoming limitations of existing methods.
• The method successfully constructs joint gene networks, enabling comparative analysis across conditions.
• Demonstrated performance on synthetic data and identified novel biological insights in medulloblastoma and glioblastoma studies.

Conclusions:

• JGNsc provides a robust framework for inferring gene regulatory networks from sparse scRNA-seq data.
• The method facilitates understanding of complex genetic architectures underlying human diseases across multiple conditions.
• JGNsc enhances biological discovery by revealing novel interactions and confirming known pathways in cancer studies.