Idea Transcript
Chapter 13.3: What could be influencing the realised response to selection? The predicted genetic response to selection is used to make a plan, to design the breeding scheme. Afterwards, genetic gain is recorded, but it often is not compared to the predictions. That is a shame, because the difference between predicted and realised selection response provides an indication of the realised success of the breeding program. It is, therefore, very wise to monitor because it provides insight in the success of the breeding program and in aspects that do not go as you initially planned them. Assumptions are not realised Why would there be a difference between estimated and realised genetic response? Consider the formula for predicting response to selection and evaluate each of the components. Were their real values the same as used in the prediction? Selection intensity To start with, the realised selection intensity may have been lower than used for predicting the response. For example, some of the animals that were selected for some reason were not able to participate in breeding. This means less superior animals will have taken their place, which will have decreased the realised genetic response to selection. Or some of the selected animals were used as parent very much more than others. This will affect the selection response. Accuracy of selection The next component is the accuracy of selection. This is influenced by the heritability and the information sources, e.g. own performance versus sib testing. The information sources are influenced quite easily. If in the prediction equation it was assumed that the EBV of all animals would be based on 5 offspring each, and in reality some animals had fewer offspring, this will reduce the accuracy. Similarly, it could also be that instead of 5, some animals had 8 offspring. This will improve the accuracy of their EBV, and thus increase the probability of selecting the genetically best animals for breeding. The heritability cannot be influenced easily. As we have seen in the chapter about genetic models, one potential way to increase a heritability is by improving the measuring method for obtaining the phenotype. The heritability may also change due to a change in additive genetic variation. Additive genetic standard deviation The next component in the formula is the additive genetic standard deviation, which is the square root of the additive genetic variation. The additive genetic variance is estimated by combining the phenotypic information and the additive genetic relationships between the animals. See the chapter on genetic models. It makes use of the fact that related animals are more alike than unrelated animals. However, if these pedigree relationships are not recorded accurately, the related animals (on paper) no longer perform that much more alike. Fewer of the similarities between the animals can be assigned to genetic relationships. Pedigree errors thus reduce the size of the estimated additive genetic variance. Even if the pedigree recording was correct and the estimated additive genetic variance is as accurate as possible, still the estimate may change somewhat across generations. As we have seen in the chapter about relationships and inbreeding, there are some forces that will have an influence on the additive genetic variation, even though the changes will not be large from generation to generation. In the longer term it does make a difference. Therefore, it is important to re-estimate the additive genetic variation at regular basis. Potential reasons for change are that selection increases the frequency of the desired alleles. Genetic drift, however, may cause the alleles that were under selection to decrease, rather than increase, in frequency. Mutations may create new variation, that cannot be predicted at forehand. Generation interval
The last component in the prediction equation is the generation interval. This only matters if you chose to express the response per year, rather than per generation. Determining the generation interval can be quite tricky, as we have seen in the chapter on genetic response to selection. It hardly ever is the same for all families, so you have to assume an average. In real life, the generation interval may be longer or shorter than anticipated, causing the realised genetic response per year to be different from the predicted one. Thus: When failing to realise predicted genetic response to selection: are all assumptions with respect to the components of the prediction equation met?