Application of Bayesian network structure learning to identify causal variant SNPs from resequencing data

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Identifying causal single-nucleotide polymorphisms (SNPs) is crucial for understanding genetic diseases.
• Bayesian network structure learning (BNSL) offers a probabilistic approach to inferring genetic relationships.

Purpose of the Study:

• To evaluate the effectiveness of BNSL in detecting true causal SNPs associated with the Affected phenotype using GAW17 data.
• To assess BNSL's performance in identifying causal variants within pre-selected candidate genes.

Main Methods:

• Applied BNSL to single-nucleotide polymorphism (SNP) genotype data from the 1000 Genomes Project pilot3 (GAW17).
• Utilized the hypergeometric distribution to assess the statistical significance of SNP subsets connected to the Affected phenotype.
• Analyzed pooled and single replicates across different ancestral populations (Asian, African, European).

Main Results:

• Exploratory analysis of pooled replicates in the Asian population sometimes identified SNP sets with more true causal SNPs than expected by chance.
• Analyses of single replicates yielded inconsistent results.
• No nominally significant results were observed in African or European populations.
• Overall, the method did not consistently identify SNP sets with a higher proportion of true causal SNPs than expected by chance.

Conclusions:

• The applied BNSL method, as implemented, is not effective for identifying causal SNPs within the GAW17 dataset, especially those that are rare.
• Further methodological refinement may be needed to improve BNSL's utility in genetic association studies with complex SNP models.