CBFI: A multi-layer gene network key node identification algorithm integrating structural-biological features and...

Area of Science:

• Systems Biology
• Bioinformatics
• Network Science

Background:

• Identifying key nodes in multi-layer gene regulatory networks (GRNs) is essential for discovering biomarkers and therapeutic targets.
• Key nodes possess both local structural advantages and global regulatory influence.

Purpose of the Study:

• To introduce the Comprehensive Biological Feature-based Importance (CBFI) algorithm for comprehensive node importance quantification in multi-layer GRNs.
• To integrate local and global network attributes for improved identification of disease-relevant nodes.

Main Methods:

• The CBFI algorithm integrates the Phenotype-Augmented Network Constraint Coefficient (PNCC) for local attributes and the Global Wandering Tenacity Coefficient (GTC) for global attributes.
• Local attributes combine structural hole theory, weak connection metrics, and gene expression-phenotype correlation.
• Global attributes utilize differential mutual information for edge weights and modified K-shell decomposition for node weights, optimized via a biased random walk model incorporating node tenacity.

Main Results:

• CBFI demonstrated superior performance in the lung squamous cell carcinoma (LUSC) multi-layer GRN (AUC=0.9346), with higher overlap with benchmark sets compared to other methods.
• The algorithm showed robustness across different ranking thresholds and adaptability in machine learning models.
• External validation on pancreatic adenocarcinoma and effectiveness in a single-layer glioma network were confirmed; CBFI_D outperformed HCIC.

Conclusions:

• CBFI offers a robust and generalizable framework for identifying biologically relevant key nodes in GRNs.
• The algorithm facilitates biomarker discovery and provides network-based therapeutic insights.