A two step Bayesian approach for genomic prediction of breeding values

Area of Science:

• Genomics
• Quantitative Genetics
• Statistical Genetics

Background:

• Genomic models with individual marker variances have limited information per parameter.
• Clustering markers with similar effects into groups with common variances offers an alternative approach.
• Marker grouping increases the degrees of freedom (df) influencing posterior variance estimation.

Purpose of the Study:

• To evaluate the effectiveness of grouping markers with similar effects for genomic prediction.
• To compare the accuracy of marker grouping methods against SNP-BLUP and Bayesian methods with marker-specific variances.

Main Methods:

• Simulated data from the 15th QTL-MAS workshop were analyzed.
• Single nucleotide polymorphism (SNP) markers were ranked by effect and grouped into sets of 150 with common variances.
• SNP-BLUP prediction models were used, with subsequent analyses focusing on subsets of markers with the largest effects.

Main Results:

• Marker grouping demonstrated higher accuracy in predicting breeding values than the standard SNP-BLUP model.
• However, the predictive accuracies achieved through marker grouping were lower than those obtained from Bayesian methods utilizing marker-specific variances.

Conclusions:

• Marker grouping offers increased power for estimating marker variances compared to individual marker variances.
• This approach is less flexible than marker-specific variance models but enhances estimation power.
• Accurate clustering of markers necessitates prior knowledge of the genetic architecture of the trait and appropriate prior parameterization.