Extension of the bayesian alphabet for genomic selection

TLDR

Milk yield and fat yield have QTL with larger effects than protein yield and somatic cell score. The study developed two Bayesian methods, BayesCπ and BayesDπ, to treat the prior probability π that a SNP has zero effect as unknown and to mitigate the impact of prior hyperparameters on BayesA and BayesB. The methods were compared by estimating the number of QTL and the accuracy of genomic estimated breeding values (GEBVs) using simulated scenarios and real data from North American Holstein bulls. BayesCπ’s π estimates are sensitive to QTL number and training size, offering insight into genetic architecture, while BayesA and BayesB do not impair GEBV accuracy; accuracies across Bayesian methods are comparable, BayesA performs well on real data, BayesCπ is computationally faster than BayesDπ, and overall BayesCπ is recommended for routine use.

Abstract

Two Bayesian methods, BayesCπ and BayesDπ, were developed for genomic prediction to address the drawback of BayesA and BayesB regarding the impact of prior hyperparameters and treat the prior probability π that a SNP has zero effect as unknown. The methods were compared in terms of inference of the number of QTL and accuracy of genomic estimated breeding values (GEBVs), using simulated scenarios and real data from North American Holstein bulls. Estimates of π from BayesCπ, in contrast to BayesDπ, were sensitive to the number of simulated QTL and training data size, and provide information about genetic architecture. Milk yield and fat yield have QTL with larger effects than protein yield and somatic cell score. The drawback of BayesA and BayesB did not impair the accuracy of GEBVs. Accuracies of alternative Bayesian methods were similar. BayesA was a good choice for GEBV with the real data. Computing time was shorter for BayesCπ than for BayesDπ, and longest for our implementation of BayesA. Collectively, accounting for computing effort, uncertainty as to the number of QTL (which affects the GEBV accuracy of alternative methods), and fundamental interest in the number of QTL underlying quantitative traits, we believe that BayesCπ has merit for routine applications.