High-Dimensional Bayesian Network Inference From Systems Genetics Data Using Genetic Node Ordering.

Area of Science:

• Genetics and Genomics
• Systems Biology
• Computational Biology

Background:

• Understanding genetic variation's impact on gene regulatory networks is crucial for explaining phenotypic variation.
• Bayesian networks are valuable for genetic mapping and modeling causal trait relationships.
• Conventional Bayesian network inference is computationally prohibitive for high-dimensional genetic data.

Purpose of the Study:

• To develop a scalable and efficient method for inferring high-quality Bayesian gene networks from large genetic datasets.
• To overcome the computational limitations of traditional Bayesian network inference methods.

Main Methods:

• Proposed a novel method involving pairwise causal inference tests to reconstruct gene node ordering.
• Inferred the Bayesian network through a series of independent variable selection problems for each gene.
• Validated the approach using simulated and real systems genetics data.

Main Results:

• The new method achieved Bayesian network likelihoods comparable or superior to conventional methods.
• Demonstrated significantly higher overlap with ground truth networks compared to existing approaches.
• Achieved computational speeds orders of magnitude faster than traditional methods.

Conclusions:

• Bayesian network inference using pairwise node ordering offers a highly efficient and scalable solution for reconstructing gene regulatory networks.
• The method provides unified false discovery rate control for rigorous network sparsity tuning and interpretation.
• This approach is particularly effective when prior information for edge inclusion is available or can be inferred.