- Accuracy of Estimated Breeding Value (EBV) is a fundamental concept in modern animal breeding because it describes how closely an EBV reflects an animal’s true, but unobservable, breeding value. Breeding value represents the animal’s underlying additive genetic merit and the genetic contribution it is expected to pass to its offspring. Because the true breeding value cannot normally be observed directly, breeders use EBVs as statistical predictions based on available information such as phenotypic records, pedigree relationships, family information, progeny performance, repeated records, correlated traits, and genomic information. The accuracy of an EBV therefore determines how confidently an animal can be ranked and selected for genetic improvement.
- Accuracy is usually expressed as the correlation between the estimated breeding value and the true breeding value. In simplified form, it can be represented as: Accuracy = Correlation(EBV, True Breeding Value). Accuracy ranges from 0 to 1, where a value close to 0 indicates little information about the animal’s true genetic merit, while a value close to 1 indicates a very strong prediction. In quantitative genetics, EBV accuracy is often denoted by r, or more specifically by r_AÂ, representing the correlation between true additive genetic value and its prediction. Accuracy should not be confused with the EBV itself: an EBV gives an estimate of genetic merit, whereas accuracy indicates how reliable that estimate is.
- A major factor affecting EBV accuracy is the amount and quality of information available for the animal and its relatives. An EBV based only on an individual’s own phenotype may have limited accuracy, particularly when the trait has low heritability. Additional information from parents, full-sibs, half-sibs, offspring, repeated records, and genetically correlated traits can substantially improve the prediction. The stronger the genetic relationship between the animal being evaluated and the individuals providing information, and the more informative and reliable those records are, the greater the potential accuracy of the EBV.
- Heritability plays an important role in determining the accuracy of selection based on individual phenotypic records. Heritability is commonly expressed as: h² = σ²_A / σ²_P, where σ²_A is additive genetic variance and σ²_P is phenotypic variance. When heritability is high, an individual’s phenotype provides relatively strong information about its additive genetic merit, so individual-based EBVs can achieve useful accuracy. When heritability is low, environmental effects contribute strongly to phenotypic differences, making the individual phenotype a weaker predictor of breeding value and increasing the importance of information from relatives, repeated records, progeny, correlated traits, or genomic data.
- The number of records and the structure of the data also influence EBV accuracy. Large datasets containing many animals, well-connected contemporary groups, repeated measurements, and reliable pedigree information generally provide better genetic evaluation. Contemporary groups are important because animals should be compared under similar environmental and management conditions. If environmental differences are not properly accounted for, the statistical model may incorrectly attribute environmental effects to genetic differences, reducing the accuracy of EBVs.
- Information from relatives is particularly valuable because relatives share genes. A pedigree relationship matrix can be used to describe expected genetic relationships among animals and incorporate this information into genetic evaluation. Parents provide information about an animal’s possible genetic background, while siblings, half-siblings, and other relatives provide additional evidence about genetic merit. However, relatives do not have identical genotypes because of Mendelian sampling, so information from several relatives is generally more informative than assuming that an individual’s genetic merit can be predicted perfectly from its parents alone.
- Progeny testing can greatly increase EBV accuracy, particularly for traits that are difficult or expensive to measure directly on selection candidates. When many offspring of a sire or dam are evaluated under appropriate conditions, their performance provides information about the parent’s transmitted genetic merit. The greater the number of informative offspring and the quality of their records, the more precisely the parent’s breeding value can be estimated. Progeny information has historically been especially important for traits such as milk production, fertility, carcass traits, disease-related traits, and other characteristics that may be sex-limited, expressed late in life, or difficult to measure directly.
- Repeated records can also increase accuracy when the same trait is measured more than once on an animal. This is particularly useful for traits with substantial temporary environmental variation. Repeatability describes the consistency of repeated phenotypic records on the same animal. When repeated records are informative and appropriately modeled, they can help distinguish persistent differences among animals from temporary environmental effects and therefore improve the prediction of genetic merit.
- Accuracy can also be increased through genetic correlation between traits. If a difficult-to-measure trait is genetically correlated with another trait that can be measured accurately and at an early age, information from the correlated trait can contribute to the EBV for the target trait. Genetic correlation can be expressed as: r_A = Cov_A(X,Y) / (σ_A,X × σ_A,Y). The usefulness of correlated information depends on the magnitude and direction of the genetic correlation, the accuracy of the information trait, and the amount of available data.
- Modern genomic selection has transformed EBV accuracy, particularly for young animals that have little or no progeny information. Genomic information provides thousands or millions of genetic markers distributed throughout the genome. These markers can be used to calculate a genomic relationship matrix and to predict Genomic Estimated Breeding Values (GEBVs). Genomic information can therefore increase the accuracy of genetic evaluation at an early age and reduce dependence on long progeny-testing periods. The improvement is particularly valuable for traits that are expensive, difficult, sex-limited, late-expressed, or have low heritability.
- The accuracy of genomic EBVs depends strongly on the quality and size of the reference population. A reference population contains animals with both genomic information and reliable phenotypic or progeny-based genetic evaluations. The closer the genetic relationship between the reference population and the animals being evaluated, and the more representative the reference population is of the target population, the greater the potential genomic prediction accuracy. Accuracy may decline when genomic prediction is applied to genetically distant populations, different breeds, substantially different environments, or populations with insufficient reference data.
- BLUP (Best Linear Unbiased Prediction) and animal-model genetic evaluation are important methods for obtaining accurate EBVs. A simplified animal model can be written as: y = Xb + Za + e, where y represents observed records, b represents fixed effects, a represents random additive genetic effects, and e represents residual effects. BLUP combines information from the animal itself and its relatives while accounting for fixed environmental effects and relationships among animals. Modern evaluations can integrate pedigree, phenotypic, repeated, progeny, and genomic information to improve prediction accuracy.
- The statistical reliability of an EBV is closely related to accuracy. A commonly used relationship is: Reliability = Accuracy². Therefore, an EBV with an accuracy of 0.80 has a reliability of approximately 0.64, while an accuracy of 0.90 corresponds to a reliability of approximately 0.81. Reliability is useful because it describes the proportion of variation in the true breeding value that is effectively supported by the prediction. However, the exact interpretation and calculation of reliability can depend on the genetic evaluation system and statistical assumptions used.
- Accuracy is particularly important when ranking animals for selection. If two animals have different EBVs but their accuracies are low, their true breeding values may differ less clearly than their EBVs suggest. As accuracy increases, breeders can have greater confidence that animals with superior EBVs genuinely possess greater additive genetic merit. This is especially important when selection decisions involve animals at a young age, when limited individual or progeny information is available.
- Accuracy also influences expected genetic gain. A simplified expression for annual genetic gain is: ΔG/year = i × r × σ_A / L, where i is selection intensity, r is selection accuracy, σ_A is the additive genetic standard deviation, and L is the generation interval. Increasing accuracy can therefore increase the expected response to selection without necessarily increasing selection intensity. Genomic selection can provide an important advantage because it may increase accuracy at a young age while simultaneously reducing the generation interval.
- The relationship between accuracy and generation interval is particularly important in modern breeding programs. Traditional progeny testing may provide highly accurate evaluations, but it can require several years before sufficient offspring information becomes available. Genomic evaluation can provide useful EBVs earlier in life, allowing animals to be selected before they have produced offspring. Earlier selection can shorten the generation interval, potentially increasing the annual rate of genetic improvement even when the absolute accuracy is not as high as that obtained after extensive progeny information.
- Accuracy must be considered separately for each trait because different traits have different heritabilities, data structures, measurement difficulties, genetic correlations, and available information. Production traits such as growth rate or milk yield may achieve relatively high accuracy when large numbers of reliable records are available. In contrast, fertility, disease resistance, survival, longevity, welfare, and some behavioural traits may be more difficult to evaluate because they often have low heritability, complex environmental influences, limited records, or later expression. For such traits, combining information from relatives, correlated traits, repeated records, and genomic data can be especially valuable.
- Multi-trait genetic evaluation can improve accuracy by analyzing several genetically correlated traits simultaneously. Instead of evaluating each trait independently, the statistical model uses information about genetic covariance among traits. This approach can be especially useful when one trait is difficult to measure but another genetically correlated trait has abundant and reliable records. However, genetic correlations can also be unfavorable, and selection based only on a highly accurate trait can unintentionally change another trait in an undesirable direction. Therefore, accuracy should always be considered together with breeding objectives, selection criteria, and the overall consequences of selection.
- Accuracy can be affected by the quality of phenotypic and pedigree data. Incorrect parentage, missing records, measurement errors, poor identification of animals, inconsistent recording systems, and inadequate adjustment for environmental effects can reduce the reliability of genetic evaluation. Good data management, accurate pedigree recording, appropriate contemporary-group definitions, standardized measurements, and well-designed breeding programs are therefore essential for obtaining reliable EBVs.
- Accuracy may also change as new information becomes available. An EBV calculated when an animal is young may initially have moderate or low accuracy because only limited information is available. As the animal produces offspring, receives additional phenotypic records, accumulates repeated measurements, or becomes connected to more relatives with records, its EBV can change and its accuracy can increase. Genomic information can provide additional evidence early in life, but genomic EBVs can also be updated as the reference population and genetic evaluation system improve.
- The genetic base of an evaluation is also important when interpreting EBVs and their accuracy. EBVs are expressed relative to a defined genetic base, and changes in the reference population or evaluation system can influence reported values. Accuracy, however, concerns the quality of the prediction rather than simply the numerical magnitude of the EBV. Breeders should therefore evaluate both the EBV and its associated accuracy or reliability when comparing animals.
- Accuracy becomes particularly important in selection for multiple traits. A breeding program may simultaneously consider growth, production, feed efficiency, fertility, health, disease resistance, survival, welfare, conformation, and adaptation. A highly accurate EBV for one trait does not automatically mean that the animal is genetically superior overall. Selection indexes can combine information from multiple traits and information sources, while the breeding objective determines the desired overall direction of genetic improvement. A simplified selection index can be written as: I = b₁x₁ + b₂x₂ + … + bₙxₙ, while a breeding objective can be represented as: H = a₁A₁ + a₂A₂ + … + aₙAₙ.
- Accuracy should also be considered in relation to genetic diversity and long-term sustainability. Selecting heavily from a small number of highly ranked animals can increase genetic concentration and potentially increase inbreeding. Therefore, breeding programs should not maximize EBV accuracy or short-term genetic gain in isolation. Inbreeding, effective population size, genetic diversity, mate allocation, and optimal contribution selection should be considered alongside EBVs to maintain a healthy and sustainable breeding population.
- The accuracy of EBVs is therefore determined by the quantity, quality, relevance, and genetic structure of the information used in genetic evaluation. Individual performance, heritability, relatives, pedigree, progeny, repeated records, correlated traits, BLUP, and genomic information can all contribute to more accurate predictions. The central objective is not simply to produce a numerical EBV, but to produce a prediction that reliably reflects the animal’s underlying additive genetic merit and supports sound selection decisions.
- In practical animal breeding, high-accuracy EBVs allow breeders to make more confident selection decisions, identify genetically superior animals at younger ages, improve the efficiency of selection programs, and increase the expected rate of genetic progress. The combination of accurate phenotypic recording, strong pedigree and population structure, advanced statistical methods, genomic information, appropriate breeding objectives, and careful management of genetic diversity provides the foundation for effective modern genetic improvement.
