Bayesian model selection for genome-wide epistatic quantitative trait loci analysis

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genetics

Background:

• Identifying epistatic quantitative trait loci (QTL) for complex traits across the genome is a significant challenge.
• The large number of potential QTL and epistatic interactions complicates genome-wide analysis.

Purpose of the Study:

• To develop a Bayesian model selection framework for identifying epistatic QTL in experimental crosses.
• To simplify calculations by restricting models to a fixed dimension using a composite model space approach.

Main Methods:

• Utilized a composite model space approach with a Bayesian framework for model selection.
• Employed indicators to specify main and epistatic effects of putative QTL.
• Developed a computationally efficient Markov chain Monte Carlo (MCMC) algorithm, incorporating Gibbs sampling and the Metropolis-Hastings algorithm.
• Incorporated prior knowledge to bound the number of detectable QTL and specify prior distributions for genetic effects.

Main Results:

• Successfully detected new epistatic QTL associated with obesity in a mouse backcross (CAST/Ei onto M16i).
• The proposed Bayesian framework and MCMC algorithm efficiently explored the posterior distribution for model selection.

Conclusions:

• The developed Bayesian model selection framework provides an efficient method for identifying epistatic QTL in complex traits.
• This approach simplifies genome-wide epistatic analysis by managing model complexity and utilizing prior knowledge effectively.