Adsa Logo White Adsa Title White

Alternative models for genetic analysis of pregnancy loss in dairy cattle.

A. Sigdel

Events

06-22-2020

Abstract:

126
Alternative models for genetic analysis of pregnancy loss in dairy cattle.
A. Sigdel*, R. S. Bisinotto, F. Pe�agaricano. University of Florida Gainesville, FL.

Pregnancy loss directly impairs reproductive performance in dairy cattle. Early pregnancy losses (before 30 d after AI) cannot be detected and are indistinguishable from conception failure. Here, we evaluated the loss of pregnancy following accurate detection of a viable embryo. As such, our objectives were to assess alternative models for genetic analysis of presence (BIN) or number (NUM) of pregnancy losses in US Holstein cows. Linear and Probit models were fitted for BIN, whereas linear and Poisson models were used for NUM. Data consisted of 14k confirmed pregnancy/abortion records on 8k Holstein cows distributed over the first 2 lactations. All models included days in milk, year-season, and types of service (insemination or embryo transfer) as fixed effects, and animal and service sire as random effects. The alternative models were compared with respect to goodness-of-fit, ranking of sires, and predictive ability in 5-fold cross-validation. Estimates of heritability ranged from 1% to 8% for BIN and 1% to 9% for NUM. Nonlinear models (Probit and Poisson) showed better goodness-of-fit than their counterpart linear models. From a breeder's perspective, an important question is whether these models yield different breeding decisions. The Spearman rank correlations between bulls' breeding values were high, from 0.84 to 0.97, suggesting a minor re-ranking. All the models exhibited similar predictive ability. Indeed, for BIN models, the mean-squared error of prediction (MSEP) ranged from 0.16 to 0.18, whereas for NUM models, MSEP values ranged from 0.14 to 0.16. Overall, our results suggest that pregnancy loss is a heritable trait, and hence, genetic selection for reduced risk of abortion is feasible. In addition, the use of nonlinear models seems a reasonable choice for analyzing pregnancy losses.

Keywords: heritability, non-linear models, reproductive performance.