Detecting genetic association through shortest paths in a bidirected graph.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWASs) typically employ marginal association tests for single-nucleotide polymorphisms (SNPs).
• These marginal tests have suboptimal power for detecting SNPs obscured by linkage disequilibrium (LD), as they assume SNP independence.
• Existing multivariate methods face computational challenges due to the large number of SNPs.

Purpose of the Study:

• To develop a novel statistical method for identifying SNPs with significant but weak marginal associations that are hidden by LD.
• To improve the power of GWASs in detecting complex genetic architectures.

Main Methods:

• A graph-based approach is proposed, constructing bidirected graphs around SNPs with moderate marginal associations (focal SNPs).
• Adjacency in the graphs is determined by LD measures.
• Shortest paths to focal SNPs are identified, and multiple regression models are fitted to SNPs within these paths to test for significance.

Main Results:

• Simulation studies demonstrate the method's superior ability to detect LD-hidden SNPs compared to marginal testing and existing multivariate techniques.
• Application to Alzheimer's Disease Neuroimaging Initiative (ADNI) data identified SNP groups near the apolipoprotein E (APOE) and semaphorin 5A (SEMA5A) genes.

Conclusions:

• The proposed graph-based method effectively detects susceptibility SNPs masked by LD.
• This approach enhances the discovery of complex genetic associations in GWASs.
• The method shows promise for identifying novel genetic loci in complex diseases like Alzheimer's.