Analyzing the effect of different approaches of penalized relationship in multistage selection schemes1

Abstract

The aim of this study was to extend optimum contribution selection to more realistic breeding schemes with multistage selection. It seems that if the last selection stage accounts for the relationship of the selected animals, then previous selection stages also account for this relationship. An extreme example was considered here: the preselection of dairy bulls that enter a progeny testing scheme. First the penalty on the average relationship in selection step 1 is assumed the same as in step 2. Thereafter, situations with different penalties on the average relationship in the 2 selection steps were analyzed. The simulation started with the generation of prior EBV, which were sampled from a truncated normal distribution. Possible candidates for further progeny testing were selected and progeny test EBV were simulated, where the progeny test was based on 100 daughters per young bull. In situations with greater accuracy of prior EBV, high trait heritability and prior EBV were available for 2,000 bulls; the results were similar for both approaches, independent of family size. However, in a situation with low accuracy of prior EBV and low trait heritability it could be observed that with increasing penalty on the average relationship, correction for relationship in stage 1 yielded in a similar genetic level compared with selecting only for high prior EBV. If the number of bulls with prior EBV increased from 2,000 to 4,000, an increasing penalty on an average relationship gave an improved genetic level. A further improvement of the results with respect to genetic level and average relationship could be observed by increasing the penalty on an average relationship in selection step 1 above that in selection step 2. Overall, this study showed that it is beneficial to use a penalty on an average relationship already for the selection of bulls that enter the progeny test. In case optimum contribution was applied with a constraint on the average relationship in stage 2, this constraint may be translated into a penalty on the average relationship, and the current results suggested that the optimal penalty in selection stage 1 should be twice that of stage 2.