CSCN: Inference of Cell-Specific Causal Networks Using Single-Cell RNA-Seq Data.

Area of Science:

• Genomics
• Computational Biology
• Systems Biology

Background:

• Gene regulation understanding is crucial for cellular activity.
• Single-cell RNA sequencing (scRNA-seq) offers cellular resolution but existing gene network inference methods overlook heterogeneity.
• Current methods like CSN and c-CSN have high false positive rates and lack causal interpretability.

Purpose of the Study:

• To develop a novel framework, Cell-Specific Causal Network (CSCN), for inferring directed, cell-specific gene regulatory relationships.
• To address limitations of existing methods by explicitly modeling causality and improving interpretability.
• To enhance the accuracy and reduce false positives in single-cell gene network inference.

Main Methods:

• CSCN framework combines causal discovery with efficient computation (kd-trees, bitmap indexing) for conditional independence testing.
• Infers sparse, interpretable causal graphs for each cell, suppressing indirect and spurious associations.
• Utilizes Causal Katz Matrix (CKM) for cell-state discrimination and trajectory analysis.

Main Results:

• CSCN achieved more accurate and stable cell-state discrimination across nine scRNA-seq datasets compared to baselines.
• CSCN representations preserved developmental structure, showing superior trajectory performance and agreement with human embryo progression.
• CSCN demonstrated the lowest false-positive rates in confounding simulations compared to CSN and c-CSN, and generalized to multiome and spatial transcriptomics.

Conclusions:

• CSCN provides a robust framework for inferring cell-specific causal gene regulatory networks.
• The method significantly improves upon existing approaches in accuracy, interpretability, and false positive reduction.
• CSCN is applicable to various single-cell omics data types, advancing the understanding of gene regulation.