- The prediction of breeding values is one of the most important processes in modern animal breeding and quantitative genetics. It helps breeders identify animals with superior genetic potential and select parents that are likely to produce improved offspring in future generations. Breeding values represent the additive genetic merit of animals for particular traits, such as milk production, growth rate, fertility, feed efficiency, disease resistance, meat quality, and longevity. By predicting these values, breeders can make more informed selection and mating decisions, improve the genetic quality of livestock populations, and achieve long-term progress in productivity, health, and adaptability.
- The fundamental principle behind breeding value prediction is that animals differ genetically, and some of these differences can be transmitted from parents to offspring. An animal’s observed performance is influenced by both its genetic makeup and its environment. For example, two dairy cows may produce different amounts of milk because of differences in inherited genetic merit, nutrition, health, housing, or management. Similarly, two beef cattle may have different growth rates because of genetic variation, feeding conditions, or disease exposure. Therefore, selecting animals based only on their observed performance can lead to inaccurate decisions. Breeding value prediction uses statistical methods to separate genetic influences from environmental effects as far as the available data and model allow.
- A breeding value is an estimate of the additive genetic contribution an animal can pass to its offspring. Additive genetic effects are particularly important in selective breeding because they describe the average effects of alleles that are transmitted across generations. An animal with a high breeding value for growth rate is expected to contribute genes associated with faster growth, on average, compared with animals having lower breeding values in the same genetic evaluation. However, the breeding value does not guarantee that every offspring will perform at the same level. Each offspring receives a random sample of genetic material from its parents, and its actual performance also depends on environmental conditions and other genetic effects.
- Breeding values are generally expressed relative to a defined genetic base or reference population. Their interpretation depends on the trait, the evaluation system, and the scale used. For example, a breeding value for milk yield may be expressed in kilograms of milk, while a value for growth may be expressed in kilograms of body weight. Some genetic evaluation systems report predicted transmitting abilities, or PTAs, instead of full breeding values. A predicted transmitting ability represents the expected average genetic contribution from a parent to its offspring and is approximately half of the parent’s breeding value under the usual additive genetic model. The exact interpretation should follow the definitions used by the relevant evaluation program.
- The accuracy of breeding value prediction depends on the quantity and quality of information available for each animal. This information may include the animal’s own performance records, records from its parents and other relatives, offspring performance, pedigree relationships, and genomic data. An animal with numerous high-quality records from itself and its relatives may receive a more accurate breeding value than an animal with little information. However, the value of additional records depends on how informative they are, whether they are collected under comparable conditions, and how strongly the trait is influenced by genetic variation.
- Heritability plays an important role in breeding value prediction. Heritability describes the proportion of phenotypic variation in a population that is attributable to additive genetic variation under the conditions studied. Traits with higher heritability generally provide more information about genetic differences through observed performance. Traits with low heritability are more strongly influenced by environmental variation or other sources of phenotypic variation, making individual performance less informative about breeding value. Fertility and many health traits often have lower heritability than some growth or carcass traits, although estimates vary among populations and recording systems. For low-heritability traits, information from relatives, repeated measurements, correlated traits, and genomic data can be particularly useful.
- One of the most widely used methods for predicting breeding values is Best Linear Unbiased Prediction, commonly known as BLUP. BLUP combines performance records, pedigree relationships, and statistical adjustments for systematic environmental effects to predict the genetic merit of animals. The method allows information from relatives to contribute to the evaluation and helps account for differences in management and production conditions. For example, if two bulls have daughters raised in different herds, a properly specified BLUP model can adjust for herd and other relevant environmental effects when comparing their genetic merit. This is more reliable than simply comparing the average performance of their daughters without adjustment.
- The animal model is a widely used framework for BLUP-based breeding value prediction. It includes the additive genetic effects of individual animals and represents relationships among them through a pedigree-based relationship matrix. Because the model uses information from multiple generations, it can estimate breeding values even for animals with limited individual records. A young animal may have no offspring records, but its parents, siblings, and other relatives may provide useful information. The animal model combines these records according to the expected genetic relationships among the animals and the assumed genetic and residual variance components.
- Sire models are another approach used in certain genetic evaluation systems. These models focus on the genetic effects associated with sires and use offspring records to estimate their genetic merit. They can be useful in populations where sire information and progeny performance are central to the breeding structure. However, a full animal model can generally represent the genetic contributions of both male and female animals more comprehensively. The appropriate model depends on the available data, the breeding system, the trait being evaluated, and the objectives of the genetic evaluation.
- Pedigree information is especially valuable when predicting breeding values. Pedigrees identify the relationships among animals and allow statistical models to account for expected genetic sharing between parents, offspring, siblings, and more distant relatives. For example, an animal may receive useful genetic information from a half sibling that has been evaluated for milk yield, growth, or disease resistance. The contribution of each relative depends on the degree of relationship, the accuracy of the relative’s evaluation, and the statistical model. Accurate pedigree records are therefore essential for reliable breeding value prediction. Missing or incorrect parentage can reduce the usefulness of family information and distort genetic relationships.
- Genomic information has significantly improved the prediction of breeding values in many livestock populations. Genomic breeding values are estimated using DNA marker information distributed across the genome. Single nucleotide polymorphisms, or SNPs, are commonly used to measure genetic variation among animals. Genomic prediction models identify statistical relationships between marker patterns and measured traits in a reference population. These relationships are then used to predict genetic merit in selection candidates. Genomic information can improve the accuracy of predictions for young animals, particularly when traditional evaluations provide limited information from individual performance or offspring records.
- Genomic best linear unbiased prediction, or GBLUP, is one of the major methods used for genomic breeding value prediction. Instead of relying exclusively on expected relationships from pedigrees, GBLUP uses a genomic relationship matrix constructed from DNA marker data. This matrix estimates genetic similarity among animals based on their observed genotypes. Genomic relationships can distinguish differences in realized genetic sharing among animals that have the same pedigree relationship. Other approaches include Bayesian regression methods and genomic prediction models that estimate the effects of genetic markers individually. The performance of these methods depends on the trait, the population’s genetic architecture, and the quality and size of the reference dataset.
- Single-step genomic BLUP, often abbreviated as ssGBLUP, combines pedigree, genomic, and phenotypic information within a unified evaluation framework. This approach allows genotyped and non-genotyped animals to be included in the same genetic evaluation. It can be especially useful in breeding programs where genomic information is available for only a subset of the population but extensive pedigree and performance records exist for other animals. By connecting these sources of information, single-step methods can improve the use of available data and support more comprehensive genetic evaluation.
- Multiple-trait breeding value prediction is another important development in quantitative genetics. Many economically important traits are genetically correlated, meaning that genetic effects influencing one trait may also influence another. For example, milk production may be associated with fertility, udder health, body condition, or longevity. Growth rate may be related to mature body size, feed efficiency, and carcass characteristics. A multiple-trait model evaluates several characteristics together and can use information from correlated traits to improve prediction accuracy. This is especially useful when a target trait is difficult or expensive to measure but another trait is more readily recorded and genetically associated with it.
- Repeated records can also improve breeding value prediction. Traits such as milk production, body weight, egg production, and feed intake may be measured multiple times during an animal’s life. These observations can provide information about both the animal’s overall genetic merit and how its performance changes over time. Random regression models can be used to evaluate traits measured across a continuous period, such as milk yield throughout lactation or body weight during growth. Such models allow genetic effects to vary with time and can provide more detailed evaluations than models that treat all repeated records as equivalent.
- The prediction of breeding values requires accurate estimates of genetic and residual variance components. Additive genetic variance describes the amount of variation in a population attributable to additive genetic differences, while residual variance represents variation not explained by the other effects included in the model. These components influence how much weight the prediction model assigns to individual records and information from relatives. Variance components can be estimated using methods such as restricted maximum likelihood, or REML, and Bayesian statistical approaches. Reliable estimates depend on suitable data, an appropriate model, and sufficient information to distinguish genetic effects from environmental influences.
- Environmental adjustments are essential for fair breeding value comparisons. Factors such as herd, year, season, sex, age, parity, feeding regime, and management group can influence recorded performance. Genetic evaluation models include relevant fixed effects to account for systematic differences and may include additional random effects when appropriate. If these influences are ignored, animals from more favorable environments may appear genetically superior even when their inherited merit is not greater. Proper adjustment allows breeders to compare animals more fairly and reduces the risk of selecting based on environmental advantages rather than genetic merit.
- Genotype-by-environment interaction can complicate breeding value prediction when animals are evaluated across different production systems. An animal with high genetic merit under intensive feeding and controlled housing may not have the same relative performance under extensive grazing or hot climatic conditions. Genetic rankings can change when the expression of genetic differences depends on the environment. Breeding programs can investigate these interactions by collecting records across environments and using appropriate statistical models. This helps identify animals that perform well in the production conditions where their offspring are expected to be raised.
- The interpretation of predicted breeding values should include consideration of their reliability. An EBV is a prediction, not a direct measurement of all the genes an animal carries. Reliability is influenced by the amount of information, trait heritability, the structure of the pedigree, genomic information, and the quality of the statistical model. Selection decisions should therefore consider both the estimated value and its uncertainty. For example, a young animal with a high EBV but limited information may be a promising candidate, but its ranking could change as additional performance records or genomic data become available. Genetic evaluation systems often update breeding values as new data are recorded.
- Breeding value prediction is closely connected to selection indexes and breeding objectives. A breeding program rarely aims to improve only one trait. Instead, breeders usually seek a balance among productivity, fertility, health, longevity, feed efficiency, product quality, and adaptability. A selection index combines information from several traits using weights that reflect their economic importance and the program’s overall goals. This helps breeders avoid excessive emphasis on one characteristic that may be genetically associated with unfavorable changes in other traits. A balanced selection index supports more sustainable improvement and better reflects the practical needs of livestock production.
- The practical application of breeding value prediction differs across livestock species. In dairy cattle, evaluations often focus on milk yield, milk composition, fertility, udder health, and longevity. In beef cattle, important traits include growth, calving ease, maternal performance, carcass quality, and reproductive efficiency. Sheep and goat breeding programs may evaluate wool quality, milk production, litter size, parasite resistance, and adaptation. Pig breeding systems often emphasize growth rate, feed efficiency, litter performance, and carcass characteristics. Poultry programs may focus on egg production, hatchability, feed conversion, growth, and disease resistance. Although the traits and recording systems vary, breeding value prediction provides a common statistical foundation for identifying genetically superior animals.
- Data quality remains one of the most important factors affecting prediction accuracy. Incorrect animal identification, missing parentage, inconsistent trait definitions, measurement errors, and incomplete environmental records can introduce bias or reduce reliability. Breeding programs should establish standardized recording procedures and implement regular quality-control checks. Modern technologies such as electronic identification, automated weighing, milk recording systems, and genomic testing can improve data collection, but their benefits depend on accurate implementation and appropriate analysis. Genetic evaluations should also be validated and updated as breeding populations, management practices, and selection objectives change.
- Artificial intelligence and machine learning are increasingly being explored to improve the prediction of animal performance and genetic merit. These methods can analyze large genomic datasets, sensor measurements, images, and complex patterns of records. In some cases, they may improve prediction by identifying relationships that are difficult to model using simpler approaches. However, predictive performance must be evaluated using independent data, and results should be compared with established quantitative genetic methods. Machine learning does not automatically replace BLUP or other genetic evaluation techniques, particularly when explicit modeling of genetic relationships, variance components, and breeding values is required.
- In conclusion, the prediction of breeding values is a fundamental process in animal breeding and quantitative genetics. It enables breeders to estimate the inherited genetic merit of animals by combining performance records, pedigree relationships, genomic information, and environmental adjustments. Methods such as BLUP, animal models, genomic prediction, and multiple-trait evaluation provide a scientific basis for selecting superior breeding animals and improving economically important traits. Reliable breeding value prediction supports faster genetic progress while helping breeders balance productivity with fertility, health, longevity, adaptability, and genetic diversity. As data collection and genomic technologies continue to advance, breeding value prediction will remain a central tool for developing efficient, resilient, and sustainable livestock populations.
