The effects of using the wrong genetic model to predict the merit of crossbred genotypes

Abstract

This paper evaluates the consequences of using the wrong genetic model when predicting the merit of previously untested crossbred genotypes. Eight models are considered, seven including different biological interpretations of two-locus epistatic interaction, plus one excluding epistatic effects. Published results from 13 genotypes generated from Hereford and Angus parental breeds were analysed, and predictions of a further seven genotypes made using different models. Under a dominance model, the predicted superiority in pregnancy rate (%) of a ¾ Hereford: ¼ Angus composite over a two-breed rotation was +1·2%, yet under all other models this was a negative value ranging from –1·9% to –3·7%. However, few such cases were found in which significant decision errors could conceivably be made. It is concluded that decisions on the choice of crossbred genotypes are generally quite robust to differences in the genetic model of the type studied here

Estimativas de efeitos genéticos em bezerros cruzados por diferentes modelos e métodos de estimação

“…Contudo, quando existem fortes relações lineares entre as variáveis independentes (multicolinearidade), a simples estimativa por quadrados mínimos dos coeficientes de regressão individuais tende a ser instável, geralmente com grande erro-padrão, e pode provocar inferências errôneas (Bergmann & Hohenboken, 1995). Kinghorn & Vercoe (1989), Cassady et al (2002) e Roso et al (2005a) reportaram problemas de multicolinearidade quando efeitos epistáticos foram adicionados ao modelo aditivo-dominante.…”

“…Contudo, quando existem fortes relações lineares entre as variáveis independentes (multicolinearidade), a simples estimativa por quadrados mínimos dos coeficientes de regressão individuais tende a ser instável, geralmente com grande erro-padrão, e pode provocar inferências errôneas (Bergmann & Hohenboken, 1995). Kinghorn & Vercoe (1989) Existem métodos alternativos de estimação que fornecem uma análise mais informativa dos dados quando existe a multicolinearidade (Chatterjee & Price, 1991;Draper & Smith, 1998). Um desses métodos é a regressão de cumeeira (Hoerl & Kennard, 1970), que consiste na adição de um coeficiente lambda à diagonal principal da matriz de correlações, visando quebrar as dependências lineares.…”

The impact of different genetic models on the optimum design of crossbreeding experiments

“…The following genotypes are considered (see also Kinghorn and Vercoe, 1989): A X (H X AH) and H X (A X HA), the two second backcrosses, as generated developing a two-breed rotation; 2BR H twobreed rotation at equilibrium, but in the phase with Hereford sires; 2BR A two-breed rotation at euqilibrium, but in the phase with Angus sires; 25 %H a two-breed unbalanced composite with 25% Hereford genes and 75% Angus genes; 75 %H a two-breed unbalanced composite -with 75% Hereford genes and 25% Angus genes. Additionally, 50%H, the balanced composite is considered, which is under the assumed genetic model equilivalent to the F 3 actually tested in l the experiment.…”

“…The group means involved in the two-breed rotation will be predicted more precisely with the parameters evolving from the experiment, including 15 genetic groups, whereas there are no considerable differences for the prediction of the composites. The balanced composite will be predicted most precisely (though not by a big margin) using the parameters from the original experiment, because more F 3 animals were (probably (Kinghorn and Vercoe, 1989;extended) Genotype classt Parental F, First backcrosses and backcrosses with F 2 dams F 2 , F, and 50%H Second rotational backcrosses Second backcrosses to one breed Two breed rotations 25%H and 75%H…”

“…It is shown that under certain circumstances the same designs are optimal for different models. The efficiency of a beef cattle experiment with crosses of Angus and Hereford cattle (Koch et al, 1985) is investigated and the effect of seven different biological interpretations of two-locus epistatic interactions (Kinghorn, 1987;Kinghorn and Vercoe, 1989) on the optimum design of this experiment is examined.…”