- Accuracy of selection describes how closely the information used to select individuals reflects their true breeding value or additive genetic merit. In quantitative genetics and animal or plant breeding, selection is effective only when individuals with genuinely superior genetic potential can be distinguished from individuals whose observed performance is high mainly because of environmental effects. Higher selection accuracy means that the individuals chosen as parents are more likely to have favorable genetic values, increasing the reliability of selection decisions and the potential for genetic gain.
- Accuracy is commonly expressed as the correlation between an individual’s true breeding value and its estimated breeding value or selection criterion. It is often written as rIHr_{IH}, where II represents the information or selection criterion and HH represents the true aggregate breeding value. The accuracy therefore ranges from 0 to 1 in many practical contexts, with values closer to 1 indicating more reliable selection. An accuracy of 0 means that the selection information provides essentially no useful prediction of genetic merit, whereas an accuracy close to 1 indicates that the information predicts genetic merit very well.
- The distinction between accuracy and related concepts is important. Selection accuracy describes how reliably genetic merit is predicted. Selection intensity describes how strongly individuals are selected, while the selection differential describes the difference between the mean phenotype of selected individuals and the population mean. Selection response is the resulting change in the population mean, and genetic gain describes genetic improvement, often measured per generation or per unit of time. These concepts interact but are not interchangeable.
- The accuracy of selection depends strongly on the type and amount of information available. Direct measurement of an individual’s own phenotype can provide information about its genetic merit, but the usefulness of that information depends on heritability. When heritability is low, phenotype contains relatively little information about additive genetic value because environmental variation makes up a large part of the observed differences. When heritability is higher, individual phenotype can provide a more informative basis for selection.
- Information from relatives can substantially improve selection accuracy, particularly for traits that are difficult, expensive, late-expressed, sex-limited, or impossible to measure directly in candidates. Records from parents, full siblings, half-siblings, progeny, or other relatives provide indirect information about an individual’s genetic merit because relatives share portions of their genetic material. The usefulness of family information depends on the degree of genetic relationship, the number and quality of records, the heritability of the trait, and the statistical model used to combine the information.
- Repeated measurements can also increase accuracy. If a trait is measured several times on the same individual, the combined records can provide a better estimate of the individual’s genetic merit than a single measurement. This is particularly relevant when environmental conditions vary between measurements. The concept of repeatability helps describe the consistency of repeated phenotypic records and the extent to which permanent individual effects contribute to observed variation.
- A major development in modern breeding is genomic selection, which uses genome-wide genetic markers to predict genetic merit. Genomic information can increase selection accuracy, especially when combined with phenotypic and pedigree information. A genomic estimated breeding value (GEBV) represents an estimate of an individual’s breeding value based partly or primarily on genomic information. The accuracy of genomic selection depends on factors such as the size and quality of the reference population, genetic relationships between the reference and selection populations, marker density, trait architecture, population structure, and the statistical prediction method.
- The relationship between accuracy and genetic gain is especially important in breeding programs. A simplified expression for genetic gain per generation can be written as ΔG≈irσA\Delta G \approx i r \sigma_A, where ii is selection intensity, rr is selection accuracy, and σA\sigma_A is the additive genetic standard deviation. This relationship shows that increasing accuracy can increase the expected genetic gain even when selection intensity remains unchanged. In practical breeding programs, however, accuracy must be considered together with generation interval, because genetic improvement per year depends not only on how accurately individuals are selected but also on how quickly generations can be produced.
- For example, suppose two breeding programs use the same selection intensity and have similar additive genetic variation. If one program has an accuracy of 0.50 and another has an accuracy of 0.80, the second program is expected to achieve greater genetic progress per generation because its selection decisions are more closely aligned with true genetic merit. If the more accurate program can also reduce the generation interval, the difference in annual genetic gain can become even larger.
- Accuracy is influenced by heritability because heritability determines how much of the phenotypic differences among individuals are attributable to genetic differences. For simple individual phenotypic selection under certain assumptions, the correlation between phenotype and breeding value is related to the square root of narrow-sense heritability. However, this relationship is not universal for all selection systems. When multiple records, relatives, pedigree information, genomic information, or other sources are combined, accuracy must be determined from the complete information structure rather than from heritability alone.
- The amount of information is another major determinant of accuracy. More informative records generally improve prediction, although the benefit of additional records depends on their quality and independence. A large number of highly related or poorly measured records may provide less additional information than a smaller number of high-quality records from informative individuals. Modern mixed models, BLUP, and genomic prediction methods are designed to combine different sources of information while accounting for relationships and environmental effects.
- The statistical model used for prediction also matters. Accurate estimation of breeding values requires appropriate modeling of fixed effects, random genetic effects, environmental effects, management groups, sex, age, location, batch, and other systematic sources of variation. Failure to account for important environmental differences can cause individuals to appear genetically superior or inferior when the difference is actually environmental. Consequently, selection accuracy is closely connected with phenotypic variance, environmental variance, genetic variance, and the design of data collection.
- The concept of accuracy is particularly important for traits that cannot be measured directly on selection candidates. For example, a breeding program may want to improve disease resistance, carcass quality, fertility, longevity, or feed efficiency. Measuring some of these traits may be expensive, require sacrificing the animal, take several years, or depend on reproductive status. Information from relatives, progeny, biomarkers, or genomic data can allow selection decisions to be made earlier and with greater accuracy.
- Accuracy is also important in multi-trait selection. Breeding programs rarely optimize a single trait. Instead, they may seek simultaneous improvement in growth, fertility, disease resistance, product quality, survival, and other economically or biologically important traits. A selection index combines information from multiple traits and sources to predict an overall breeding objective. The accuracy of selection then depends on how well the index predicts the desired aggregate genetic merit and on the genetic covariance and genetic correlation among traits.
- Genetic correlations can create both opportunities and limitations. If two traits have a favorable genetic correlation, improving one trait may also improve the other. If they have an unfavorable correlation, selection for one trait may produce an undesirable correlated response in another. High accuracy in predicting one trait does not automatically guarantee accurate prediction of an overall breeding objective. The quality of the information used for each trait and the genetic relationships among traits must be considered.
- Genotype–environment interaction (G×E) can also influence the meaning of selection accuracy. An individual’s genetic performance may differ across environments, and a breeding value estimated in one environment may not perfectly predict performance in another. In such situations, selection accuracy should be evaluated in relation to the target population and production environment. Incorporating records from relevant environments can improve the usefulness of selection decisions when genetic rankings change across environments.
- Accuracy should also be distinguished from repeatability and heritability. Repeatability describes the consistency of repeated measurements, while heritability describes the proportion of phenotypic variance attributable to genetic variance under a particular population and environment. Accuracy, by contrast, concerns the relationship between available selection information and genetic merit. These quantities are related, but none of them should be treated as interchangeable.
- In genomic selection, accuracy can change over time as the relationship between the reference population and the selection candidates changes. Prediction accuracy may decline when candidates become genetically distant from the population used to train the prediction model. Regularly updating the reference population with new phenotypic and genomic records can help maintain prediction performance. Population size, genetic diversity, marker effects, trait architecture, and the persistence of linkage disequilibrium between markers and causal variants all influence genomic prediction accuracy.
- High selection accuracy can improve breeding efficiency, but maximizing accuracy is not always the only objective. Collecting additional phenotypes may require substantial time and expense. Waiting for progeny records may increase accuracy but also increase the generation interval. Genomic selection can sometimes provide an attractive compromise by allowing relatively young individuals to be evaluated before extensive phenotypic or progeny information is available. The optimal breeding strategy therefore balances accuracy, selection intensity, generation interval, cost, genetic diversity, and long-term breeding objectives.
- Selection accuracy also matters for maintaining genetic diversity. Extremely strong selection based on highly accurate information can rapidly increase genetic gain, but if only a small number of individuals are chosen as parents, it may also reduce the effective population size and increase inbreeding. Modern breeding programs therefore often consider accuracy together with relationship information and constraints designed to manage inbreeding and preserve useful genetic variation.
- In natural populations, the same general principle applies. Natural selection changes allele frequencies when phenotypic or behavioral differences affect survival and reproduction. Selection is more efficient when phenotypic differences reliably reflect genetic differences associated with fitness. Environmental variation, developmental plasticity, genetic correlations, and genotype–environment interaction can all influence the relationship between observed phenotype and inherited genetic merit.
- Accuracy of selection is therefore a central concept linking quantitative genetics, breeding value, heritability, selection intensity, selection response, genetic gain, and genomic selection. Reliable selection decisions require information that predicts genetic merit rather than simply identifying individuals with favorable observed phenotypes. As breeding programs increasingly combine pedigree, phenotypic, genomic, and environmental information, improving selection accuracy has become one of the major ways to increase the efficiency and predictability of genetic improvement.
- Understanding accuracy also provides a foundation for interpreting modern breeding technologies. BLUP, genomic prediction, GEBVs, genomic selection, selection indices, and multi-trait evaluation can all be viewed as methods for improving or exploiting information about genetic merit. When accurate predictions are combined with appropriate selection intensity and a suitable generation interval, breeding programs can achieve faster and more sustainable genetic gain while maintaining control over inbreeding and genetic diversity.
