A Multi-Omics Data Integration Framework for Gene Regulatory Network Inference Based on Contrastive Learning.

Area of Science:

• Computational Biology
• Systems Biology
• Bioinformatics

Background:

• Gene Regulatory Networks (GRNs) are crucial for maintaining cellular stability and function during differentiation.
• Existing methods struggle to effectively integrate multi-omics data for cell-type-specific GRN inference.

Purpose of the Study:

• To introduce CLMOGRI, a novel multi-omics framework for inferring transcription factor-gene regulatory networks.
• To enhance the integration of diverse omics data for more accurate GRN construction.

Main Methods:

• CLMOGRI utilizes heterogeneous networks and contrastive learning to embed multi-omics data into a unified feature space.
• Random walk techniques are employed for feature extraction and node similarity measurement.
• Contrastive learning is applied to predict gene regulatory relationships.

Main Results:

• CLMOGRI significantly outperforms existing baseline methods in Area Under the Precision-Recall Curve (AUPR) and F-Score.
• The framework effectively captures multi-omics information for robust GRN inference.
• CLMOGRI successfully identifies critical nodes and modules within GRNs, enhancing interpretability.

Conclusions:

• CLMOGRI offers a powerful and interpretable approach for multi-omics GRN inference.
• The framework improves our understanding of complex biological interactions and cellular regulation.
• CLMOGRI advances the utility of GRNs in biological systems analysis.