Comparative Research on Breeding Value Prediction for Milk Yield in Cattle

The paper compares four methods for breeding value prediction for milk production in cattle. The research was performed in experimental farm of the National Research Development Institute for Animal Biology and Nutrition (IBNA), Balotesti, Ilfov, Romania. The biological material consisted of 190 animals, 86 with performances, animals belonging to Montbeliarde, Black Spotted Romanian and Brown breeds. The studied trait was the test day milk yield. To enter the statistical analysis animals must fulfil the following requirements: minimum three test-days, minimum 100 days of lactation and a maximum of 330 days of lactation. The animal's breeding value was estimated separately based on the following biometric models: Wood, Ali and Schaeffer, Legendre Polynomial and P-Spline. The degree of agreement between the rankings of the same individuals using the four models was quantified with Spearman's rank correlation. Results showed a high degree of similarity between the 4 biometric models, leading to the conclusion that any of them could be used to predict animal breeding value. Although the results are similar, we recommend, for the practical genetic evaluation of cattle, for the specific trait (milk yield), the use of P-Spline function because its description of the lactation curve evolution was closest to the biological dynamics.