- Genomic prediction is a statistical approach used to estimate the genetic merit of animals by combining genome-wide genetic marker information with phenotypic records and other relevant genetic data. It is a central component of modern genomic selection, enabling breeders to identify promising animals at an early age and improve the efficiency of livestock breeding programs. Genomic prediction is widely applied or investigated in cattle, pigs, poultry, sheep, goats, and other animal populations to improve traits such as growth, milk production, feed efficiency, fertility, disease resistance, longevity, and environmental adaptation.
- Traditional genetic evaluation relies on observed performance, pedigree information, and records from relatives or offspring to estimate an animal’s breeding value. Although these methods remain important, they may require many years of data collection for traits expressed late in life or measured only in one sex. Genomic prediction uses information from genetic markers distributed across the genome to estimate genetic merit before all relevant performance or progeny records become available. This can shorten the generation interval, improve selection decisions, and potentially increase the rate of genetic gain per year.
- The central objective of genomic prediction is to estimate an animal’s genomic estimated breeding value (GEBV). A GEBV represents a prediction of the animal’s additive genetic merit relative to a defined genetic evaluation base. It is not a direct measurement of genetic value but an estimate derived from genomic data, statistical models, and information from a suitable reference population. These predictions help breeders rank selection candidates according to their expected genetic contribution to future generations.
- Genomic prediction depends on a reference population containing animals with genomic information and reliable phenotypic records or appropriate genetic evaluation data. These animals provide the training information needed to establish relationships between genome-wide markers and genetic merit. Statistical models use the reference population to estimate marker effects or genomic relationships, then apply the resulting model to genotyped animals whose breeding values need to be predicted. The quality, size, genetic composition, and relevance of the reference population strongly influence prediction accuracy.
- The genomic information used in prediction commonly comes from single-nucleotide polymorphisms (SNPs) measured using genotyping arrays or derived from sequencing data. Before analysis, researchers perform genomic data quality control to identify unreliable samples, missing genotypes, inconsistent animal identities, and low-quality markers. Phenotypic records also require careful checking, with appropriate consideration of factors such as age, sex, herd, feeding, management, and environmental conditions. Poor-quality genomic or phenotypic data can reduce prediction accuracy and introduce bias into genetic evaluations.
- Several statistical methods are used for genomic prediction. One widely adopted approach is genomic best linear unbiased prediction (GBLUP), which uses a genomic relationship matrix to represent genetic similarity among animals. Bayesian methods estimate marker effects under different assumptions about the distribution of genetic effects, while other approaches may use haplotypes, sequence variants, or machine-learning techniques when these provide reliable predictive information. The choice of model depends on the trait, genetic architecture, population structure, data availability, and objectives of the breeding program.
- In GBLUP, the genomic relationship matrix is a key component because it estimates how genetically similar animals are based on their observed marker information. This differs from a pedigree-based relationship matrix, which estimates expected relationships from recorded ancestry. Genomic relationships can reflect the actual combinations of genetic material inherited by animals and may help improve predictions when pedigree information is incomplete. Other prediction methods may model marker effects directly, but they still depend on appropriate data preparation and statistical assumptions.
- Genomic prediction is particularly valuable for polygenic traits, which are influenced by many genetic variants, often with small individual effects. Examples include milk yield, growth rate, feed conversion, fertility, longevity, and several disease-resistance characteristics. Instead of relying exclusively on a few markers identified through genome-wide association studies (GWAS), genomic prediction combines information across many genomic regions to estimate overall genetic merit. This allows it to capture contributions from numerous variants that may not individually reach stringent statistical significance in association studies.
- The accuracy of genomic prediction depends on several factors, including the size and quality of the reference population, trait heritability, genetic relationships between training animals and selection candidates, marker information, and the statistical model used. Traits with low heritability or unreliable phenotypic records may require larger and more informative reference populations. Prediction accuracy may also decrease when the model is applied to genetically distant populations, different breeds, or later generations that are less closely related to the original reference animals.
- Genomic prediction accuracy is commonly assessed by comparing predictions with reliable independent information about genetic merit, such as later-life records, progeny information, or suitable genetic evaluations. Validation methods must be designed carefully to avoid using information in the testing data that has already influenced model training. For example, when the objective is to predict young animals from future generations, validation should reflect the genetic distance between those animals and the reference population. Reliable validation helps breeders determine whether a prediction model is suitable for practical selection decisions.
- One of the main advantages of genomic prediction is that it enables earlier genetic evaluation. Young animals can be genotyped and ranked before they have produced offspring or expressed traits that require years to measure. In dairy cattle, for example, genomic predictions can help identify promising bulls and heifers before extensive progeny testing is completed. Similar strategies can be used in other livestock species when suitable genomic data and reference populations are available. Earlier selection can reduce generation intervals and increase annual genetic improvement.
- Genomic prediction is also useful for traits that are difficult, expensive, or impractical to measure on every breeding candidate. These may include feed efficiency, methane emissions, carcass composition, disease resistance, resilience, and certain welfare-related traits. Once reliable relationships between genomic information and these traits have been established in a reference population, predictions can be generated for unmeasured candidates. However, the reliability of such predictions depends on the relevance and quality of the training data and the consistency of trait definitions across populations.
- Another important application is multi-trait genomic prediction, in which information from several traits is analyzed jointly. Some traits are genetically correlated, meaning that information about one trait can help predict another. For example, a well-recorded trait may provide useful information about a difficult-to-measure characteristic if their genetic relationship is sufficiently strong and stable. Multi-trait prediction can improve accuracy, but the benefit depends on the strength of genetic correlations, the quality of the available records, and the structure of the population.
- Genomic prediction can support broader breeding objectives through selection indexes that combine predictions for several traits. Livestock improvement programs often need to balance production, fertility, health, survival, adaptation, and animal welfare. Selection based on a single predicted trait may produce unwanted changes in other characteristics because of genetic correlations. Combining genomic predictions with clearly defined economic or strategic breeding weights helps breeders select animals that better match the overall breeding objective.
- Genomic prediction can also contribute to managing genetic diversity and inbreeding. Genome-wide marker data provide information about genetic relationships and patterns of homozygosity, which can help breeders design mating plans and avoid excessive use of closely related animals. When combined with strategies such as optimal contribution selection, genomic prediction can help balance short-term genetic gain with the retention of genetic variation. This is important for maintaining future selection potential and reducing the risk of inbreeding depression.
- Despite its benefits, genomic prediction has several limitations. Establishing and maintaining a reliable reference population requires investment in genotyping, phenotyping, data management, and statistical analysis. Predictions may be less reliable when data are limited, records are inaccurate, genetic relationships are weak, or the target population differs substantially from the training population. Prediction models may also need updating as new generations emerge, breeding objectives change, and additional phenotypic records become available. Genomic prediction should therefore be supported by continued monitoring and validation rather than treated as a permanently fixed evaluation system.
- Genomic prediction does not replace phenotypic recording or conventional genetic evaluation. Reliable performance records remain essential for developing reference populations, updating prediction models, evaluating genetic trends, and monitoring whether selection objectives are being achieved. Genomic predictions estimate genetic potential, while actual performance continues to depend on nutrition, health, management, environment, and other non-genetic factors. A high GEBV therefore does not guarantee a particular performance level under every production system.
- In practical animal breeding, genomic prediction is most effective when integrated with accurate phenotypic records, appropriate statistical models, clear breeding objectives, and responsible management of genetic diversity. Breeding organizations should monitor prediction accuracy, update reference populations, and assess whether predictions remain useful across generations and environments. They should also consider health, fertility, welfare, and long-term adaptability rather than focusing exclusively on short-term production gains.
- Overall, genomic prediction is a fundamental tool for estimating genetic merit in modern animal breeding. By combining genome-wide marker information with reliable phenotypic and genetic evaluation data, it supports earlier selection, improved breeding value estimation, and potentially faster genetic progress. Its effectiveness depends on the quality of the reference population, the suitability of the statistical model, and careful validation. When used responsibly, genomic prediction can contribute to productive, healthy, genetically diverse, and sustainable livestock populations.
