Bayesian modeling of haplotype effects in multiparent populations.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Quantitative trait loci (QTL) analysis in multiparental populations like the Collaborative Cross (CC) and Heterogeneous Stocks (HS) is crucial for understanding genetic architecture.
• Accurate estimation of founder haplotype and diplotype effects is challenged by uncertainty in haplotype composition, small sample sizes, and complex genetic backgrounds.

Purpose of the Study:

• To introduce Diploffect, a general Bayesian model for estimating founder haplotype effects at QTL in multiparental populations.
• To provide a coherent framework for estimating haplotype and diplotype effects, incorporating uncertainty, dominance, genetic background, and complex data structures.

Main Methods:

• Developed a general Bayesian model, Diploffect, utilizing probabilistic haplotype reconstruction as prior information.
• Employed two computational approaches: Markov chain Monte Carlo (MCMC) sampling and importance sampling with integrated nested Laplace approximations.
• Simulated QTL data in incipient CC and Northport HS populations to compare Diploffect with approximations and Haley-Knott regression.

Main Results:

• Diploffect provides accurate posterior distributions for haplotype effects and composition at QTL.
• The model effectively handles uncertainty in haplotype composition, sample size, and data structure.
• Simulations demonstrated Diploffect's accuracy and trade-offs compared to Haley-Knott regression and its own approximations.

Conclusions:

• Diploffect offers a robust framework for estimating genetic effects in complex multiparental populations.
• The model's ability to incorporate various sources of uncertainty enhances the precision of QTL effect estimation.
• Diploffect provides valuable insights into phenotype variation influenced by diplotype substitutions at modeled loci.