Single-cell causal network inferred by cross-mapping entropy.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding cellular functions and states.
• Inferring causal relationships within GRNs is challenging due to data noise and nonlinear molecular dynamics.

Purpose of the Study:

• To introduce a novel causal criterion, neighbor cross-mapping entropy (NME), for robust GRN inference.
• To evaluate NME's performance on both simulated and real-world single-cell RNA sequencing (scRNA-seq) data.
• To demonstrate NME's utility in identifying cell states and improving downstream analyses.

Main Methods:

• Developed the neighbor cross-mapping entropy (NME) criterion to quantify continuous causality between variables.
• Applied NME to steady-state and time-series data, including scRNA-seq datasets.
• Compared NME's performance against existing GRN inference methods using benchmark datasets.

Main Results:

• NME demonstrated superior performance in inferring GRNs compared to existing methods on benchmark datasets.
• NME successfully inferred cell-type-specific GRNs and identified distinct cell states from scRNA-seq data.
• GRNs inferred by NME improved the accuracy of single-cell clustering and other downstream analyses.

Conclusions:

• NME offers a powerful and accurate approach for inferring causal gene regulatory networks from complex biological data.
• The method facilitates the discovery of novel cell types/states and the prediction of cell type-specific network modules.
• NME's foundation in continuous causality provides a robust framework for analyzing gene interactions in scRNA-seq data.