Genomic breeding value prediction and QTL mapping of QTLMAS2010 data using Bayesian Methods

Area of Science:

• Quantitative genetics
• Genomic prediction
• Statistical genomics

Background:

• Bayesian methods, including BayesA and BayesB, are used for genomic breeding value prediction (GEBV) with high-density single nucleotide polymorphisms (SNPs).
• A modified Bayesian method, BayesCπ, was developed to estimate the proportion of SNPs with zero effects (π) and a common variance for non-zero effects.
• GEBV accuracy is typically assessed by correlating predictions with observed phenotypes.

Purpose of the Study:

• To predict GEBVs using the BayesCπ method on the QTLMAS2010 dataset.
• To compare the accuracy of BayesCπ with other methods like BayesB and best linear unbiased prediction (BLUP).
• To identify quantitative trait loci (QTLs) and evaluate their additive effects.

Main Methods:

• Application of the BayesCπ method for GEBV prediction.
• Comparison with BayesB (varying π values) and BLUP (using genomic or numerator relationship matrices).
• QTL detection using GEBV variances in 10-SNP windows and a novel significance thresholding approach based on pedigree-simulated genotypes.

Main Results:

• BayesCπ achieved the highest GEBV accuracy, comparable to BayesB with π=0.99.
• BayesB accuracy decreased as π decreased; polygenic effects had minor impacts on accuracy and bias.
• A 10% chromosome-wise threshold identified 15 QTLs, while a 20% threshold identified 21 QTLs, with few false positives.

Conclusions:

• The BayesCπ method, without polygenic effects, was optimal for the QTLMAS2010 data, showing superior accuracy and minimal bias.
• The 10-SNP window variance approach effectively detected over half of the true QTLs with a low false positive rate.
• This study highlights the utility of BayesCπ and a novel QTL significance criterion in genomic prediction and QTL mapping.