NSIF-FS8
Larry Young, U.S. Meat Animal Research Center Clay Center, Nebraska
George Cammack, Nebraska seedstock producer.
Every record on a trait that is measured on an animal is made up of two components: genetic and environmental. The genetic component is strictly a function of the genes possessed by the animal. Selection, which creates genetic improvement, operates only on the genetic component and not the environmental component that is always involved with the observed measurement. The environmental component is not passed from parent to progeny and, therefore, needs to be accounted for when determining the genetic value of an animal. Some of these environmental factors can be accounted for mathematically; i.e., parity of the sow or sex of the piglet. However, other factors such as health, feed, and management can affect a pig's performance but cannot be accounted for very easily. These are referred to as unknown sources of environmental variation.
The best method to account for unknown environmental effects is to compare animals that are subjected to the same environment. If a group of animals that are expected to respond similarly to these unknown effects are raised together, they then form what is known as a contemporary group. A contemporary group is defined as a uniformly managed group of animals of similar breed composition, age, and sex. An example might be all Duroc male piglets farrowed within 30 days of each other that are weaned and grown together in the same pen. These piglets are similar in age and identical in sex and breed and, therefore, we would expect them to respond to unknown effects similarly. They are also raised together, thereby having the same effects of feed, weather, health, etc. The concept of a contemporary group as a means of controlling nongenetic effects is basic to animal breeding and genetic improvement.
By comparing the performance of each individual animal in the contemporary group to the average of its contemporaries we can determine a more precise estimate of the genetic merit of each individual. Comparisons are most easily made by two common methods: deviations and ratios. Deviations are simply the difference between the performance of an individual and the mean performance of its contemporaries.
The ratio method simply involves dividing the animal's own performance by the average performance of its contemporaries and multiplying that result by 100. For example, consider a boar that gains at a rate of 2.0 lb./day while his contemporaries gained at an average of 1.5 lb./day. His deviation would be +0.5 lb./day (2.0-1.5 lb.), and his ratio would be (2.0+1.5) x 100 = 133. The average deviation is zero and the average ratio is always 100. Therefore, this boar gained 0.5 lb./day or 33% (133-100) faster than his contemporaries. The advantage of the deviation is that it is expressed in the units of measure being used, but one must know the average value of the trait and the variation of the deviations from the average value. The advantage of the ratio is that we do not have to remember the variability associated with any trait, and the mean ratio is always 100. A ratio of 125 means 25% greater than their contemporaries no matter what trait we are talking about.
Deviations and ratios should be used only to compare animals within a contemporary group because different contemporary groups could be comprised of various numbers of animals or subjected to different environmental conditions. Furthermore, the genetic ability of the pigs contained in two different contemporary groups is probably not identical. Growth rate deviations in two contemporary groups from different herds may mean something very different. Consider the following example for postweaning growth rate:
Deviation
from Genetic
contem- mean of Actual
porary contem- observed
group porary growth
Boar average group rate
A +1.0 0.9 1.9
B +0.5 1.8 2.3
If you used the boar with the largest deviation across contemporary groups you would choose Boar A. However, because all the boars in the second contemporary group had better genetic ability for growth rate the better boar is really Boar B. Even though he had a deviation of only + .5, his genetic potential for growth is better than that of Boar A since the genetic mean of boar B's contemporary group was higher than Boar A's (1.8 vs. 0.9 lb./day). Therefore, deviation and ratios should be used to select only pigs within their own contemporary group. Comparisons across contemporary groups using deviations or ratios could lead to improper selections, which could slow genetic progress.
The concept of true breeding value has been discussed in other fact sheets in this series. A pig's true breeding value is never really known for any trait because it is not directly measurable. Therefore, some means of predicting or estimating breeding values for traits are required. Predictions of breeding values are known as Estimated Breeding Values (EBVs) and result from the application of genetic theory to performance records.
Several methods that depend upon the available information can be used to calculate EBVs. All methods are based upon the concept of the deviation of an individual's or relative's records from their respective contemporary group averages and multiplying the deviation by a weighting factor. In the simplest case where only one record on the individual pig is available the formula is:
EBV = b (P-P) where b is the weighting factor; in this example, b is the heritability (h to the 2nd power) or percent of the variation in the trait under genetic control P is the individual's record and P is the average of the individual's contemporary group.
An example would be as follows: Boar A has an adjusted backfat measurement of .65 inches and his contemporaries average .75 inches. Given the h2 of backfat thickness as 0.4, then Boar A has an EBV for backfat thickness of 0.4 (.65 - .75) = -.04 inches. This means that genetically he is leaner than his contemporaries by .04 inches.
As more information (full-sib, half-sib, progeny data, etc.) becomes available, the formula can be extended to include the additional data from those sources as shown:
This is known as a selection index because it includes information from several sources used in estimating the genetic merit of an individual pig. The values of the weighting factors (b's) change depending upon several factors including the number of records, sources of information, heritability, pedigree relationships, and genetic correlations. However, in the case of the relatives, knowing the average performance of groups of relatives, i.e., full-sibs, half-sibs, etc. is sufficient. For traits such as number of pigs farrowed, there could be four or five records on each sow. Utilizing as many records as possible on each pig will also improve our ability to estimate breeding values. The sources of information contributing records to the estimate must be known. Records would be from full-sib, half-sib, and progeny or from correlated traits on any of those groups. Generally any increase in the heritability or amount of information, such as using more sources of information (half-sibs, full-sibs, etc.), larger groups, more records per individual or from correlated traits, will increase the accuracy of the estimated breeding value.
The heritability of a trait is very important in estimation of breeding values. Heritability (h to the 2nd power) is simply defined as that portion of the observable variation in a trait that is due to genetic control or breeding value. An indication of the relationship between an animal's own performance as a deviation from its contemporaries and its breeding value for a given trait is the heritability. Therefore, it provides an estimate of how much of an animal's superiority or inferiority we expect will be passed on to its progeny. A trait with a high heritability means that an animal's own performance is a good estimate of its breeding value for the trait. With lowly heritable traits the individual performance and breeding value are poorly related and the performance of the individual will not be a very accurate indicator of its breeding value.
Pedigree relationships become important when records are available on relatives. Relationships measure the correlation between breeding values resulting from the likeness in pedigree of two individuals. Full-sib and half-sib records would contribute much more information to predicting an individual's EBV than records from a second cousin. The more closely two animals are related, the more Important the records become as a predictor of breeding value of one individual for the other.
Genetic correlations are useful because information on one trait can be used in predicting the breeding value for another trait. Genetic correlations allow the use of records on another trait as a further source of information that will improve the estimate of the breeding value compared to that estimated without information from a genetically correlated trait.
An EBV is an estimate of the pig's genetic value for that specific trait and can give an indication of the expected progeny performance. Estimated breeding values can be expressed as ratios or deviations where the average EBV in a population is zero (expressed as a deviation) or 100 expressed as a ratio. Both methods are commonly used, so be certain of the method before making any selection decisions.
The total estimated genetic value of an animal is called an EBV, but a pig transmits only a sample one-half of its genetic value (genes) to each of its progeny. The Expected Progeny Difference (EPD) is a commonly used name that represents this one-half sample of a pig's genetic value that is passed to its offspring. Therefore, an EPD is simply one-half the EBV. The average EPD in a population is zero, and values are expressed as deviations (+ or -) from the average. EPDs could be expressed as ratios, but the current convention is the deviation format. The concept of an EPD is very easy to understand because it is truly the expected progeny difference in performance. If a boar has an EPD of +0.3 lb./day for growth rate, we expect his progeny to grow 0.3 lb./day faster than progeny of average boars. Another very useful feature of EPD is that it is comparable between two animals that are evaluated in the same contemporary group or groups linked together by relative information. For example: If the EPDs for backfat thickness for two boars, A and B, are -0.1 and -0.3 inches, respectively, progeny of boar B would be expected to have 0.2 inches less backfat than progeny of boar A. Therefore, EPDs can be very useful in not only estimating the animal's genetic worth as a parent but also in comparing two potential candidates for selection as replacement breeding stock.
Predicting the expected performance of progeny from planned matings is one use of EBVs and EPDs. However, remember the difference between an EBV and EPD so that expected progeny performance can be calculated correctly. Since an animal transmits only one-half of its breeding value, this must be taken into account. The following example will help to illustrate this point: Suppose a boar has an EBV of -6 days to 230 lb. and his mate has an EBV of -2 days. The expected performance of their progeny will be A days to 230 lb. calculated as one-half the sire plus dams estimated breeding values (-6 + -2)/2. Remember that an animal's EPD for a given trait is one-half its EBV for that trait since the animal passes on only a sample one-half of its breeding value. Therefore, the same boar would have an EPD of -3 days (-6 x 1/2) and his mate would have an EPD of -1 day (-2 x 1/2). Since EPDs already account for the one-half contribution from each parent, the expected progeny performance is simply the sum of the sire plus dam EPDs. In this example the expected progeny performance is A days (-3 + -1). Therefore, it is important to know whether you are using EBV or EPD as an evaluation of genetic worth.
Accuracy values (ACC) are associated with EBVs and EPDs and give an idea of reliability or certainty of the estimate. The ACC figures range from zero to one, with higher values representing more accuracy or reliability. An EBV with an accuracy of less than .40 is not very reliable and is probably subject to change as more information is collected. An accuracy value of .95 represents a very reliable estimate of breeding value and one would expect very little change as more data are added. This suggests something else about an ACC value that is very important. Generally, low accuracy values are caused by very few pieces of information on the animal being evaluated or by low heritability for a given trait. As more information is added to the data set, perhaps by records on full-sibs, half-sibs, and progeny, the accuracy or reliability of the estimate increases. So an accuracy value gives an indication of how much information was used in calculating the EBV or EPD. Low ACC values suggest very little information was available, while larger values point to several pieces of information being used in the calculation of the estimate of breeding value.
Possible Change (PC) is another common way of expressing accuracy or reliability of an estimate of genetic merit. Possible changes are generally used only with EPDs because they are easier to interpret with estimates expressed as deviations. A PC is expressed as a +/- value. For instance, a boar may have an EPD of -12 +/- 3 days to 230 lb. The PC gives an upper and lower limit or range to the PPD, and we would expect his true progeny difference to be within this range 67% of the time. In this example the EPD is -12 days but we would be correct 2 out of 3 times (67%) if we guessed that his true progeny difference was -9 (-12 + 3) to -15 (-12 - 3) days. A nice feature of possible change values is that if 67% is not a high enough level of confidence, you can increase your chances to 95% by simply doubling the PC value. In this case we would be correct 95 times out of 100 (95%) if we guessed that the true progeny difference was in the range of -6 (-12 + 6) to -18 (-12 - 6) days. Possible change values emphasize the fact that EPDs will change over time especially if a lot of new information is added. The smaller the PC value the more accurate and reliable is the estimate of progeny difference. This is opposite to the ACC values, where a higher value represents more reliability. We must know which measure of reliability we are dealing with because of this difference.
Estimates of genetic merit, whether they are EBVs or EPDs, provide a very useful tool for breeders. They are the best available estimates of an animal's genetic merit. The progress and success of a selection program is based upon the ability to estimate genetic merit accurately and use these estimates in selection decisions.
RR 8/94
Cooperative Extension work in Agriculture and Home Economics, State of Indiana, Purdue University and U.S. Department of Agriculture cooperating: H.A. Wadsworth, Director, West Lafayette, IN. Issued in furtherance of the acts of May 8 and June 30, 1914. The Cooperative Extension Service of Purdue University is an equal opportunity/equal access institution.