Single-step Bayesian regression methods for genomic evaluation of milk yield of Murrah buffaloes.

Abstract

In this Research Communication we describe the application of single-step Bayesian regression (ssBR) models to predict milk yield of Murrah buffaloes. Milk production records of 2,026 cows in their first lactation were used. Using 270-day cumulative milk yield records as phenotype, genomic breeding values were predicted and their accuracies and dispersions were calculated by five methods: BayesA (ssBA), BayesB (ssBB), BayesC (ssBC); Bayesian Lasso (ssBL); and Bayesian ridge regression (ssBRR). For models based on mixture distributions (ssBB and ssBC), the proportions of markers having effect (π) were assumed as fixed, with respective values of 99% or 90%, or as unknown, where two approaches to estimate π were applied (ssBayesBπ and ssBayesCπ). The accuracy values found ranged from 0.550 (ssBBπ) to 0.584 (ssBCπ) and, the dispersion estimates ranged from 0.867 (ssBA) to 0.958 (ssBRR). The results indicated that Bayesian Lasso was the most suitable model for genetic evaluation of milk yield by buffaloes, considering accuracy and dispersion as criteria.