Bayesian estimation of genetic parameters for superovulatory response traits in Japanese Black donor cows using count data models

Abstract

Listen

Translate article

We estimated genetic parameters for two in vivo embryo production-related superovulatory response traits—total number of embryos and oocytes (TNE) and number of good embryos (NGE)—in Japanese Black donor cows through Bayesian count regression analysis. We used 20,257 records of superovulation treatments from 1546 Japanese Black cows, with 1102 (5.4%) zero-count records for TNE and 3533 (17.4%) for NGE. Two generalized mixed linear models (MLMs; repeatability animal models)—Poisson (POI) and zero-inflated Poisson (ZIP) regression models—were fitted to the untransformed phenotypic records. A Gaussian MLM was also fitted to untransformed phenotypic records (GAU), natural log-transformed records (LOG), and records with Anscombe's variance stabilizing transformation (ANS). The estimated heritabilities and repeatabilities of TNE were 0.30 and 0.43 by POI, 0.35 and 0.47 by ZIP, 0.27 and 0.36 by GAU, 0.21 and 0.31 by LOG, and 0.24 and 0.35 by ANS, respectively. Those of NGE were 0.29 and 0.36 by POI, 0.31 and 0.40 by ZIP, 0.18 and 0.25 by GAU, 0.19 and 0.24 by LOG, and 0.20 and 0.25 by ANS, respectively. Under the ZIP, the estimated heritabilities and repeatabilities of the probability of zero counts were 0.43 and 0.71 for TNE and 0.42 and 0.51 for NGE, respectively, and the rank correlations between estimated breeding values of the 1546 donor cows for superovulation response and those for the probability of zero count were around −0.40 for TNE and −0.50 for NGE.