Across-country genomic prediction of bull fertility in Jersey dairy cattle

Abstract

The use of information across populations is an attractive approach to increase the accuracy of genomic predictions for numerically small breeds and traits that are time-consuming and difficult to measure, such as male fertility in cattle. This study was conducted to evaluate genomic prediction of Jersey bull fertility using an across-country reference population combining records from the United States and Australia. The data set consisted of 1,570 US Jersey bulls with sire conception rate (SCR) records, 603 Australian Jersey bulls with semen fertility value (SFV) records and SNP genotypes for roughly 90,000 loci. Both SCR and SFV are evaluations of service sire fertility based on cow field data, and both are intended as phenotypic evaluations because the estimates include genetic and nongenetic effects. Within-and across-country genomic predictions were evaluated using univariate and bivariate genomic best linear unbiased prediction models. Predictive ability was assessed in 5-fold cross-validation using the correlation between observed and predicted fertility values and mean squared error of prediction. Within-country genomic predictions exhibited predictive correlations of around 0.28 and 0.02 for the United States and Australia, respectively. The Australian Jersey population is genetically diverse and small in size, so careful selection of the reference population by including only closely related animals (e.g., excluding New Zealand bulls, which is a less-related population) increased the predictive correlations up to 0.20. Notably, the use of bivariate models fitting all US Jersey records and the optimized Australian population resulted in predictive correlations around of 0.24 for SFV values, which is a relative increase in predictive ability of 20%. Conversely, for predicting SCR values, the use of an across-country reference population did not outperform the standard approach using pure US Jersey reference data set. Our findings indicate that genomic prediction of male fertility in dairy cattle is feasible, and the use of an acrosscountry reference population would be beneficial when local populations are small and genetically diverse.

Use of insemination data for joint evaluation of male and female fertility in predominantly seasonal-calving dairy herds

“…Similarly, the predic-Haile-Mariam and Pryce: JOINT EVALUATION OF MALE AND FEMALE FERTILITY tion of masked future data is more accurate using model 1b than model 1a by up to 33%. That bull solutions predicted from the model fitted with the genetic and PE bulls as one term improved the accuracy prediction of bull solutions for NRR, compared with a model that ignored the genetic effect, is in general agreement with several recent studies (e.g., Abdollahi-Arpanahi et al, 2017;Rezende et al, 2020).…”

“…DataGene has been using these data to calculate SFV, which is a phenotypic evaluation of AI bulls for fertility of their semen (Carrick et al, 2000). The SFV for Jersey bulls from Australia and the service bull conception rate from the United States were jointly used for genomic evaluation of male fertility recently by Rezende et al (2020). The NRR of bulls collected between 2004 and 2019 were used for this study.…”

“…This low accuracy could be because the evaluation relies on the male side of the pedigree, which could be incomplete for some bulls. An alternative is to use genomic-enabled evaluation (e.g., Abdollahi-Arpanahi, et al, 2017;Rezende et al, 2020), as most bulls have genotype data. Genomic-enabled predictions of bull fertility so far are based on 2-step genomic evaluation models using the service sire conception rate or SFV as response variable (e.g., Abdollahi-Arpanahi et al, 2017;Rezende et al, 2020).…”

“…An alternative is to use genomic-enabled evaluation (e.g., Abdollahi-Arpanahi, et al, 2017;Rezende et al, 2020), as most bulls have genotype data. Genomic-enabled predictions of bull fertility so far are based on 2-step genomic evaluation models using the service sire conception rate or SFV as response variable (e.g., Abdollahi-Arpanahi et al, 2017;Rezende et al, 2020). It would be interesting to assess the accuracy of genomic prediction from models that use the raw NRR and genotype data in a single-step genomic prediction model.…”

“…With the recent shift in the dairy industry toward the intensive use of young genomically selected bulls, exploring causes of variation in bull fertility early has become an emerging area of research (Taylor et al, 2018;Haile-Mariam and Pryce 2019). The renewed interest in assessing the extent of genetic variation in male fertility is partly because of the potential to carry out genomic-enabled screening of bulls before extensive use for semen collection and mating (Abdollahi-Arpanahi et al, 2017;Rezende et al, 2020).…”