(new) Chapter 8.2: In more detail: breeding value estimation

In the overview you have seen that there are several ways to rank the animals. General aim in animal breeding is to rank the animals as well as possible. A tool for ranking the animals is the estimated breeding value. The more accurate the estimate of the breeding value, the better results of the subsequent breeding can be expected. We will now look into the three methods of estimating breeding values that were described in the overview in more detail.

But first we need to know what an estimated breeding value really is. How do we get from information on the phenotypes of the animals and their genetic relationships (pedigree) to an estimate of the breeding value of the animals? In animal breeding we use the principle of regression to achieve this. In figure 1 you can see this principle visualised. If we would plot the true breeding value on the y-axis against the phenotypic superiority on the x-axis, then we can calculate a regression line through the data points. In real life, unfortunately, we cannot create such a plot as we don’t know the true breeding values. Instead, we try to find the regression coefficient that, in combination with the phenotypic superiority, would best predict the genetic superiority or true breeding value (TBV). The art of estimating breeding values is based on finding ways to come to the best regression coefficient. This immediately also highlights a critical point in breeding value estimation: it is a linear regression coefficient but animals with the same phenotypic superiority do not always have the same genetic superiority. For some animals, like the animal indicated with a circle in the figure, the TBV is very different from the EBV, whereas for others the EBV would be the perfect estimate of the true breeding value. Part of this difference in how well the EBV resembles the TBV is caused by the fact that the phenotype can be influenced quite substantially by the environment. Therefore, simultaneous to finding the best regression coefficient, it is also important to try to make the phenotypic superiorities fit the regression line as well as possible. In the rest of this chapter we will discuss some options to work on both these problems: predict the best regression coefficient, and make the phenotypic superiority fit the regression line as well as possible.