Using the Pareto principle in genome-wide breeding value estimation.

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Bioinformatics

Background:

• Genome-wide breeding value (GWEBV) estimation relies on marker effect distributions.
• Traditional methods like GBLUP use constant variance normal priors, while Bayesian methods offer flexibility but are computationally intensive.
• Existing Bayesian methods often struggle with computational demands due to Markov chain Monte Carlo sampling.

Purpose of the Study:

• To develop a computationally efficient Bayesian method for GWEBV estimation.
• To incorporate the Pareto principle to model SNP effects with varying importance.
• To introduce MixP, a novel method utilizing a mixture of normal distributions for SNP effects.

Main Methods:

• Applied the Pareto principle to weight marker loci based on their contribution to genetic variance.
• Developed MixP, a Bayesian method with a mixture of two normal priors (large and small variance) for SNP effects.
• Utilized iterative equation solving, reducing computational load significantly compared to MCMC methods.

Main Results:

• MixP demonstrated accuracy comparable to or exceeding existing methods in simulations.
• The method simplifies hyper-parameter tuning by reducing them from two to one.
• Computational efficiency was improved by two orders of magnitude, making it suitable for large datasets.

Conclusions:

• MixP offers a computationally feasible and accurate Bayesian approach for GWEBV estimation.
• The method effectively models the distribution of SNP effects, aligning with the Pareto principle.
• MixP is well-suited for the analysis of high-density marker and whole-genome sequence data in animal breeding.