Comparison of methodologies for predicting breeding values using simulated data

Revista Brasileira de Zootecnia

Torres, Robledo de Almeida²  Silva, Martinho de Almeida e¹  Regazzi, Adair José²  Lopes, Paulo Sávio²  Braccini Neto, José³  Euclydes, Ricardo Frederico² 

[1] Universidade Federal de Minas Gerais;Universidade Federal de Viçosa, Viçosa;Universidade Federal do Rio Grande do Sul, Porto Alegre

关键词: animal model;    GenSys;    mixed-model;    sire model    IntroduçãoA maioria dos dados disponíveis aos melhoristas da espécie animal não apresenta os requerimentos de amostragem aleatória;    ou seja;    os dados são provenientes de experimentos de seleção ou de rebanhos que estão sob seleção (Henderson;    1973;    Meyer & Thompson;    1984;    Gianola et al;   

DOI  :  10.1590/S1516-35982004000700006

学科分类：兽医学

来源: Universidade Federal de Vicosa

【 摘 要 】

Seven populations with different genetic means and different data structures resulting from replacing 0, 10, 25, 50, 75, 90 and 100% of sires simulated by GENESYS program were used to compare the mixed-model methodology (MMM) with the GenSys methodology, according to the accuracies and predicted error variances (PEV) of the breeding values, estimated under a sire or an animal model. Under the sire model, lower accuracy and greater PEV for both methodologies were observed as selection intensity increased, due to the reduced accuracy of estimation of the fixed effects and the reduced average number of progenies per sire. The superiority of MMM under the sire model was due to increase in the accuracy of the estimated fixed effects. Under the animal model, the tendencies were different. For the MMM, accuracy increased and PEV remained practically constant, as selection intensity increased, whereas for the GenSys methodology, reduced accuracy and an increase on the PEV as selection intensity increased were observed. Greater accuracies for MMM were due to the increased number of genetic ties, whereas the reduced accuracy for GenSys was due to the lower accuracy of the estimation of the fixed effects. The superiority of MMM was a result of the greater accuracy of the levels of fixed effect estimates and the use of the full relationship matrix, which included a base-population of unrelated, non-endogamic and non-selected individuals. The MMM was superior to GenSys according to the accuracy and PEV, for this structure of simulated selected data, mainly under the animal model.

【 授权许可】

CC BY-NC