EstratÃgias de prediÃÃo de cruzamentos nÃo realizados em anÃlise dialÃlica, utilizando informaÃÃes de marcadores moleculares. / Alternatives for predicting effects of missing crosses in diallels, with information from molecular markers.

AUTOR(ES)
DATA DE PUBLICAÇÃO

2004

RESUMO

Diallel crosses are largely used by breeders to assist choosing parents and/or populations. One constraint of such designs is the effort required to obtain crosses, the amount of which grows exponentially as the number of parents increases. This work aimed at evaluation some approaches to predict effects of missing crosses in diallel analysis, comprising 120 crosses among 22 common bean parents (from four different races), carried out in three sites of Minas Gerais State in Brazil. A matrix of 0âs and 1âs was available, with respect to 136 RAPD loci, aiming the application of methodologies with different principles to predict missing crosses. Such methodologies were: (i) an additive model, where prediction in made with the general combining abilities of the parents (molecular information not used); (ii) a distance model, similar to (i), but with additional covariate, corresponding to the genetic distance (calculate with molecular data) among parents of each cross; (iii) factor regression models (four variants were considered) using covariates that were subsets of the first four principal components relative to the matrix of 0âs and 1âs; (iv) a mixed models (best linear unbiased prediction) with two variants, considering and non-considering the inbreeding that arose in crosses where parents were related. In each model, losses of 12, 24, 36, 48 and 60 crosses were simulated one thousand times in each situation for the additive, the genetic distance and factor models. For the mixed models, in each situation 250 simulations were performed. The effects of such crosses were predicted considering the remaining data. In each simulated condition, Pearsonâs and Spearmanâs coefficient of correlation were estimated among predictions and observed phenotypic values of missing crosses. Across simulations, averages, maximum and minimum values, and standard deviations of those coefficients were calculated. With all models, there was a trend of increasing the coefficients of correlation with less missing crosses. Factor models showed the highest coefficients of correlation among predicted and observed values. Best linear unbiased prediction and the distance model showed similar results. Inclusion of a covariate in the mixed model, to account for inbreeding, brought no further improvement on the predictions. This may have been a result of the quite similar values of inbreeding estimated in the set of crosses.

ASSUNTO(S)

melhoramento vegetal crossing marcador molecular molecular markers common bean feijÃo cruzamento

ACESSO AO ARTIGO

Documentos Relacionados