Bayesian network prior: network analysis of biological data using external knowledge.

Area of Science:

• Computational Biology
• Bioinformatics
• Systems Biology

Background:

• Reverse engineering gene interaction (GI) networks from experimental data is complex due to network intricacy and data noise.
• Integrating external biological knowledge is crucial for improving the accuracy of GI network construction.

Purpose of the Study:

• To propose a novel framework for learning GI networks from experimental data.
• To incorporate external biological knowledge using a Bayesian Network Prior (BNP) within the structure learning process.

Main Methods:

• Utilized Bayesian networks (BNs) for learning GI networks from experimental data.
• Developed a Bayesian Network Prior (BNP) to represent relationships between evidence types for gene interactions.
• Employed BNP to calculate candidate graph probabilities during structure learning.

Main Results:

• The proposed approach accurately identifies underlying interaction networks across synthetic, simulated, and real biological data.
• Demonstrated high accuracy even when prior information was distorted.
• Outperformed existing methods in GI network reconstruction.

Conclusions:

• The framework effectively integrates external biological knowledge into GI network inference.
• The method provides a robust and accurate solution for reverse engineering complex biological networks.
• The BNP approach enhances the reliability of gene interaction network construction from diverse data sources.