Recovering gene regulatory networks in single-cell multi-omics data with PRISM-GRN

Area of Science:

• Genomics
• Systems Biology
• Computational Biology

Background:

• Gene regulatory networks (GRNs) are essential for understanding cellular functions, development, and diseases.
• Existing GRN inference methods often lack the ability to integrate multi-omics data and prior biological knowledge effectively.
• Current approaches struggle to provide biologically interpretable insights into gene regulation.

Purpose of the Study:

• To develop PRISM-GRN, a novel Bayesian model for reconstructing cell type-specific GRNs.
• To integrate single-cell RNA sequencing (scRNA-seq) and single-cell assay for transposase-accessible chromatin sequencing (scATAC-seq) data with prior GRN information.
• To enable precise, robust, and interpretable inference of causal GRNs.

Main Methods:

• PRISM-GRN utilizes a Bayesian probabilistic framework to combine scRNA-seq, scATAC-seq, and known GRN data.
• The model employs a biologically interpretable architecture based on transcription factor (TF) expression and chromatin accessibility.
• It uses a mechanism-informed generation process and a prior-GRN-primed inference process for GRN reconstruction.

Main Results:

• PRISM-GRN demonstrated superior performance in GRN reconstruction compared to seven baseline methods across four benchmarking datasets.
• The model achieved higher precision in sparse regulatory interaction scenarios and captured causality in gene regulation.
• PRISM-GRN showed flexibility in handling unpaired omics data and limited prior GRN information.

Conclusions:

• PRISM-GRN offers a novel paradigm for precise, robust, and interpretable causal GRN exploration.
• The model's ability to integrate multi-omics data and prior knowledge enhances the understanding of cellular heterogeneity and disease mechanisms.
• PRISM-GRN has significant potential for identifying cell type-specific or context-specific GRNs in various biological research applications.