Genome-Wide Association Analyses Based on Broadly Different Specifications for Prior Distributions, Genomic Windows,...

Area of Science:

• Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Genome-Wide Association (GWA) studies are crucial for identifying genetic markers linked to traits.
• Current methods like EMMAX treat marker effects differently, potentially limiting statistical coherence.
• Alternative prior distributions offer a more unified statistical framework for GWA analysis.

Purpose of the Study:

• To compare the performance of different prior distributions (Gaussian, BayesA, SSVS) in GWA analysis.
• To investigate the impact of quantitative trait loci (QTL) effect skewness and number on GWA results.
• To evaluate the utility of adaptive genomic windowing strategies for GWA studies.

Main Methods:

• A simulation study was conducted using 43,266 SNP genotypes from 922 pigs.
• Three prior categories were compared: Gaussian, BayesA (heavy-tailed), and SSVS (mixture of Gaussians).
• Genomic regions were defined by single SNPs, fixed-size windows, or adaptively determined windows based on linkage disequilibrium.

Main Results:

• Stochastic search and variable selection (SSVS) and BayesA priors demonstrated superior receiver operating curve properties.
• Approximate maximum a posteriori (MAP) approaches for BayesA and SSVS were computationally feasible but yielded unpromising results.
• Adaptive genomic window lengths were found to be advantageous for GWA studies.

Conclusions:

• SSVS and BayesA priors provide a more robust framework for GWA analysis compared to traditional methods.
• Adaptive windowing strategies enhance the effectiveness of GWA studies.
• MAP-based inferences for complex models in GWA analysis require careful consideration due to sensitivity to starting values.