Genetic prediction of feed efficiency in South African Holstein cattle.

Abstract

Feed efficiency is a trait of outstanding importance in dairy cattle; however, it is difficult and expensive to measure and, therefore, not easy to improve through selection. The current study investigated the possibility of predicting the genetic merit for feed efficiency, in South African Holstein cattle, using routinely recorded easy-to-measure traits. The first two objectives were mainly to develop and validate models to predict dry matter intake (DMI) and gross feed efficiency (GFE) using milk production traits and live weight (LW). Data consisted of 30 daily measurements of DMI, milk yield (MY), energy-corrected milk (ECM), butterfat yield (BFY), protein yield (PROY), lactose yield (LACY), butterfat percent (BFP), protein percent (PROP), lactose percent (LACP), and 25 daily LW records of a group of 100 first-parity Holstein cows, fed a total mixed ration. Similar measurements were also collected from a group of 110 multiparous Holstein cows, in lactations 2 to 6. Gross feed efficiency was calculated as kg ECM divided by kg DMI. Forward stepwise regression analyses were performed to develop the models, using the PROC REG procedure of the Statistical Analysis System (SAS) software. Within-herd validation of the models for robustness and accuracy, was subsequently conducted by performing regression analyses between actual and predicted DMI and GFE records. The developed models reliably predicted daily DMI (kg/day) and GFE from milk, butterfat yield and/or live weight, with accuracies ranging from 66 to 98%. Validation of the best model to predict DMI, for first-parity cows, yielded a fairly moderate R2 value (0.49) and a low root mean square error (RMSE) (1.46 kg/day), while the best prediction model for GFE yielded a fairly high R2 value (0.64) and a low RMSE (0.13). The best prediction model for GFE, for multiparous cows, had a fairly moderate R2 value (0.54) and a low RMSE (0.06), upon validation, suggesting reasonable robustness and accuracy. The developed models, therefore, present an opportunity to easily generate large quantities of phenotypic data on individual cow DMI and GFE, at a relatively low cost, which can be used to achieve accurate selection for feed efficiency.

The third objective was, essentially, to assess the extent of genetic variability for the predicted traits, in order to evaluate their utility as selection criteria for feed efficiency. First, repeatability animal models were used to estimate genetic parameters for predicted gross feed efficiency (pGFE) and its relationship with energy-corrected milk (ECM) in the first three parities, using the ASReml software. Data of 11,068 test-day milk production records on 1,575 Holstein cows that calved between 2009 and 2019, were used. Predicted gross feed efficiency was calculated using the models developed in the first two objectives. Heritability estimates for pGFE ranged from 0.09 ± 0.04 in mid lactation to 0.18 ± 0.05 in late lactation. Estimates were moderate for primiparous (0.21 ± 0.05) and low for multiparous (0.10 ± 0.04) cows. Repeatability and heritability estimates across all lactations were 0.37 ± 0.03 and 0.14 ± 0.03, respectively. Genetic correlations between pGFE in different stages of lactation ranged from 0.87 ± 0.24 (early and mid) to 0.97 ± 0.28 (early and late), whereas a strong genetic correlation (0.90 ± 0.03) was obtained between pGFE and ECM, across all lactations. The average genetic merit for pGFE, across all lactations, increased at a marginal rate of 0.0058 per year, for cows born during the period 2007 to 2017. The low to moderate heritability estimates for pGFE suggest potential for genetic improvement of the trait through selection, albeit with a modest accuracy of selection. The high genetic correlation of pGFE with ECM may, however, assist to improve accuracy of selection for feed efficiency by including both traits in multi-trait analyses. Due to the scarcity of information on the genetic variation exhibited by predicted dry matter intake (pDMI) from milk components, further genetic analyses were undertaken. Such analyses were important to determine whether pDMI could be a useful selection criterion for feed efficiency in dairy cattle. These analyses under the third objective involved estimation of heritabilities, repeatabilities and genetic correlations among predicted dry matter intake and gross feed efficiency, by repeatability animal models. Data consisted of 440,062 test-day records of 62,695 cows, in the first three parities, that calved between 2009 and 2019. Predicted dry matter intake was generated from milk yield data, using the developed prediction model, and pGFE was derived as kg ECM divided by kg pDMI. Heritability estimates ranged from 0.05 ± 0.02 for pGFE in mid lactation to 0.13 ± 0.03 for pDMI in late lactation. Estimates of heritability across parities were 0.08 ± 0.02 and 0.13 ± 0.02 for pGFE and pDMI, respectively. Corresponding estimates of repeatability across parities were 0.26 ± 0.01 and 0.41 ± 0.01 for pGFE and pDMI, respectively. Genetic correlations between pDMI and pGFE were moderate and negative in early (-0.42 ± 0.24) and mid lactation (-0.20 ± 0.24), and low and positive in late lactation (0.05 ± 0.17). The genetic correlation between pDMI and pGFE decreased with increase in parity, from 0.26 ± 0.16 to -0.09 ± 0.17. The low heritability estimates for pDMI and pGFE indicate low accuracy of selection for these traits in South African Holstein cattle. This can, however, be improved through multi-trait analyses including traits with which they are correlated.

Results of the current research pave the way for achieving genetic improvement in feed efficiency in the South African Holstein cattle population. This can go a long way towards the development of a more profitable and environmentally sustainable dairy industry. Higher rates of genetic change can be attained through genomic selection, by using the predicted phenotypes to identify genes or markers associated with feed efficiency.