Mining genetic epidemiology data with Bayesian networks I: Bayesian networks and example application (plasma apoE...

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Analyzing genotype-to-phenotype relationships is challenged by vast single nucleotide polymorphism (SNP) data.
• Data mining methods are crucial for addressing these analytical complexities.
• Belief (Bayesian) networks are introduced as a novel approach for these analyses.

Purpose of the Study:

• To introduce and demonstrate the application of Belief networks for genotype-to-phenotype relationship studies.
• To analyze the association between SNPs in the apolipoprotein E gene and plasma apoE levels.
• To showcase the capability of Belief networks in handling complex biological data.

Main Methods:

• Utilized Belief networks, a graphical probabilistic model, to represent joint multivariate probability distributions and conditional independences.
• Employed heuristic search algorithms and scoring criteria for optimal network topology estimation.
• Applied Bayesian methods and bootstrapping to evaluate the statistical significance of edge strengths.

Main Results:

• Applied Belief networks to 20 SNPs in the apolipoprotein E gene and plasma apoE levels in 702 individuals.
• Identified a strong association (edge) between SNP 4075 (apolipoprotein E epsilon2 allele) and plasma apoE level.
• Demonstrated that Belief networks can effectively model complex, uncertain biological processes.

Conclusions:

• Belief networks provide an effective framework for genotype-to-phenotype analysis.
• The method can learn from data and integrate prior biological knowledge.
• This approach offers a robust solution for analyzing complex genetic data and uncovering genotype-phenotype links.