Gene regulatory network inference based on causal discovery integrating with graph neural network

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene regulatory network (GRN) inference is crucial for understanding biological systems.
• Causality-inspired methods offer a more rational approach than correlation-based methods for inferring regulatory relationships.

Purpose of the Study:

• To propose GRINCD, a novel GRN inference framework.
• To integrate graph representation learning and causal asymmetric learning for robust GRN construction.
• To consider both linear and non-linear regulatory relationships.

Main Methods:

• Utilizing graph neural networks to generate high-quality gene representations.
• Applying the additive noise model for predicting causal regulation between gene pairs.
• Designing and assembling dual-channel architecture for enhanced prediction accuracy.

Main Results:

• GRINCD demonstrated superior or comparable performance against state-of-the-art methods across diverse datasets.
• The framework effectively infers gene regulatory relationships, considering complex interactions.
• Experimental results validated the robustness and accuracy of the proposed GRINCD framework.

Conclusions:

• GRINCD offers a novel and effective approach for gene regulatory network inference.
• The framework shows promise in identifying key factors involved in cancer development.
• This work contributes to advancing the field of GRN construction and biological systems analysis.