Will R. Getz – Fort Valley State University

Introduction

Predicting genetic merit or breeding value, involves the use of data from several sources including:

• Information on an individual’s own phenotype.
• Information on the genotype or performance of ancestors and(or) collateral relatives of that individual.
• Information on the genotype or performance of descendants, including progeny.

Records from any of the three sources of information may impact on the particular trait of importance. Information can be used from associated genetically-related or genetically correlated traits. For example if reducing the level of dystocia (difficult birth) is part of the selection plan, then consideration of birth weight records may assist in predicting trouble ahead of time. The two traits tend to be genetically-related. That is, they are influenced by some of the same genes, genes on the same chromosomes, the same genes effecting more than one trait. Birth weight may also be an indication of fetal size, in which a bigger fetus is more likely to result in more dystocia (difficult birth).

There two technologies that are commonly used in genetic prediction. We will not discuss them in depth because that would require a background in mathematics and statistics beyond the scope of this training tool. However we want users of this training course to have a general appreciation of when the technologies should be used and what they are capable of doing when various kinds of records are kept and utilized.

Selection index

Selection index theory was developed as a method for genetic prediction and as a means of combining traits in order to select animals in an economically optimal manner. In this section we will focus on use of the selection index as a predictor of merit or of breeding value. An index is a combination of various kinds of phenotypic information and appropriate weighting factors. Usually the weighting factors reflect some economic value associated with the given phenotypic information.

The data used in a selection index comes from the several sources mentioned above — performance, pedigree and progeny data. That data will become available at various phases of life. Unborn or very young animals have only pedigree information or that from ancestors and perhaps some collateral relatives. As animals get older they acquire performance data of their own. For sex-limited traits, e.g. milk production, they may have no data on their own performance. If the animals are selected to become parents, they will generate progeny data, and if they are reasonable popular they will have large amounts of progeny data.

In a selection index, each item of phenotypic information is normally expressed as a deviation from a contemporary group mean. This accounts for environmental differences between contemporary groups. A simple example of an index follows the following form:

The calculations involved are simple and require only a hand calculator. More sophisticated indexes can be handled by a typical personal computer and spreadsheet software. The kinds of indexes used in most livestock improvement programs require the computing capacity of modern computers.

Selection indexes may be restricted in application on the farm or ranch, but they are very useful nonetheless. Suppose you are breeding meat goats and want to compare your bucks on the basis of progeny performance within your herd of meat goats. You could use a selection index to produce progeny-based “Estimated Breeding Values” or “Expected Progeny Difference” for each buck. Sometimes predictions of genetic merit or breeding value will be less than the actual data suggest. This is because mathematically, those individuals with less information will have more conservative (closer to the mean value) outcomes. As the amount of information increases, the closer the collected data should reflect the true breeding value.

There are several factors affecting accuracy of prediction when using selection indexes:

• Number of records
• Heritability
• Repeatability
• Pedigree relationship

As heritability increases, so does accuracy of prediction, regardless of the source of information, individual, half siblings, or progeny. Accuracy also increases with pedigree relationship. On the average the individual in question will have 25% of its genes in common with half brothers or sisters (usually through a common sire), and 50% of its genes in common with progeny. Accuracy of prediction increases with the number of records. Clearly, more records provide more information on which to base predicted breeding value.

When heritability is high, the individual’s own record is especially valuable. When heritability is low, then the value of performance records on relatives become much more valuable and necessary for prediction. Progeny records are the ultimate source of information for predicting the value of an individual as a parent. A key issue to know is that with enough progeny, accuracy in predicting breeding value is high even when heritability is low.

Why? Because progeny records provide a measure of the value of the genes that an individual transmits. With enough progeny the environmental effects, the influence of breeding values of mates, and the effects of Mendelian sampling tend to even out over this large number of progeny. The average performance of many progeny is a good indication of an individual’s breeding value.

As mentioned previously, meat goat breeders assume that it is very difficult to make genetic change in traits that are lowly heritable, because of the difficulty of identifying genetically superior bucks or does. This is true if you depend solely on individual performance records, but with a large number of progeny records, the problem of low heritability can be overcome. Those animals with the better breeding values can be accurately identified.

Best linear unbiased prediction

The selection index is a powerful method for genetic prediction. However, a basic assumption is that the performance information used comes from genetically similar contemporary groups. On different farms or ranches or from different periods of time this assumption can not be made. Best linear unbiased prediction (BLUP), which is an extension of selection index theory, is designed for this type of situation, i.e. where data is available from genetically dissimilar groups. Much of the statistical theory behind BLUP was worked out in the 1960’s, but it was not until the 1980’s that computers and mathematical algorithms had advanced enough for widespread use of this tool. Like selection indexes, it requires the simultaneous solution of a number of equations reflecting the multiple sources of performance information used.

BLUP is the preferred method for large-scale genetic evaluations, typically those undertaken by breed associations. The performance records used in such evaluations usually come from field data, and represent a multitude of contemporary groups. The statistical model most often used has the capability of utilizing all animals in a population in the process of predicting the breeding value of potential breeding stock. The model accounts for differences in the mean breeding values of contemporary. They account for the fact that superior performance in a genetically inferior contemporary group is not the equivalent of superior performance in a genetically superior contemporary group.

All traits have what is called a direct genetic component, the effect of an individual’s genes on its performance. However, some of those traits have a maternal component which is the effect of genes in the dam that influence the performance of the individual through the “environment” provided by the dam. Goat weaning weight would be an example of a trait with both direct and maternal components. BLUP procedures are capable of separating the direct and maternal components of a trait, and will provide genetic predictions for both. To be complete in this discussion, it should be mentioned that the paternal effect is a third component. Fertility measurements that are considered traits of the dam or offspring but are affected by a male’s fertility and physical ability to breed, are said to have a paternal component. An example would be conception rate.

Genetic evaluation

Genetic evaluation represents the application of genetic prediction technologies provided by selection indexes and BLUP. The purpose is to allow genetic comparison of animals in different herds. This is useful because you may believe your senior herd buck is outstanding based on his individual performance, and from progeny records in your herd. However, without some mechanism for comparing him with sires owned by other breeders, you have no objective way of knowing how good he really is when compared to others in the breed as a whole.

Facilities at San Angelo State University, Langston University, Fort Valley State University, Western Kentucky University, and Pennsylvania State University have been set up to do this on a centralized scale. Private initiatives have also been developed where groups of breeders work together on their own farms and ranches. These efforts enable breeders to select individuals for breeding from a larger pool of candidates. Rather than be limited to the animals they own themselves, breeders can select from a much larger population, e.g., an entire breed.

Larger-scale genetic evaluation speeds the rate of genetic change by increasing accuracy of prediction. When records from an entire breed are used for prediction, accuracy of prediction increases by virtue of the sheer volume of information. The first step in making this kind of selection tool available is to get the records. Breeders and associated commercial producers must keep and collect records of performance for a numbers of years. Traits to be included in the performance pool will necessarily need to be identified. In the meat goat sector there are an increasing number of people keeping performance records on commercially important traits. The absolute number of breeders doing this remains small however. Sire summaries are composed of lists of genetic predictions, accuracy values, and other useful information.

Some of the first of this type information for meat animals grew out of statewide Record of Performance programs across the country. Dairy Herd Improvement (DHI) programs were the first in large animals, and even before them the poultry industry, with assistance from USDA, collected data and issued progeny test results in statewide programs. Eventually the private sector took over sire evaluation in the poultry industry as developing and crossing inbred lines became the breeding technique of choice. In large animals individual breeds undertook performance testing programs which involved data from individual breeders across the nation. Genetic predictions of choice in sire summaries today are expected progeny differences or EPD’s. In the dairy field, including dairy goats, the comparable data summaries are published in the form of estimated transmitting abilities or ETA’s.

The history of genetic evaluation is filled with periods of strong opposition in the early stages. Breeders are not quite ready to collect and utilize individual performance data on traits that are commercially important. Some meat goat associations are now in the early stages of collecting performance data, and beginning the long task of accumulating enough accurate data for meaningful comparative summaries.

Major problem areas in genetic evaluation include faulty data — pedigrees, performance records, adjustment factors, and errors in contemporary group assignment. Other challenge areas include lack of relationship among contemporary groups; and genotype by environmental interaction. These interactions occur when the difference in performance between two or more genotypes changes from environment to environment.

Correlated responses to selection

Selection for one trait rarely affects just that one trait. Usually other traits are affected as well. Genetic change in one or more traits resulting from selection of another is termed correlated response to selection. An example might be the correlation between rate of growth and feed utilization efficiency. It is probably that selection for growth or weight for age will frequently cause an improvement in feed utilization efficiency in meat goats.

The major cause of correlated response is a fifty dollar word called pleiotropy. Pleiotropic effects occur when a gene influences more than one trait. Another cause of correlated response is linkage. If major genes affecting two traits are located in the same chromosome, there is a strong tendency for them to be inherited together. Selection for one trait increases the frequency of alleles positively influencing that trait and, at the same time, increases the frequency of linked alleles as well. Linked genes do not stay together for ever. Sooner or later recombination breaks the linkage, and so linkage must be considered a temporary cause of correlated responses.

Similar polygenic traits are probably influenced by many genes with pleiotropic effects. Growth traits are an example. Some genes effect growth rate during only certain periods of life, while others affect growth in general and through several periods of life. Selection for one of these causes a correlated response in the others.

There are times when it is wise to select for a correlated trait rather than to select directly for a trait of interest — to use indirect selection rather than direct selection. Some traits are too expensive or difficult to measure directly. An example might be feed utilization efficiency, which requires individual feeding and record keeping that can be relatively expensive. Another reason to select for a correlated trait is that accuracy of selection may be greater for the correlated trait than for the trait of interest. For example measuring a buck’s scrotal circumstance is an indicator of fertility in the buck and age at puberty in related does.

The down side to correlated responses is that if we are unaware, or choose to ignore, unfavorable genetic correlation, selection for one trait can lead to undesirable responses in others. Total selection for growth rate and size may lead to difficulties with dystocia which can have dire economic consequences.