Briefly describe the various approaches for the estimation of genomic estimated breeding values. Which of these approaches is in common use and why?

Genomic estimated breeding values (GEBVs) are estimates of an individual's genetic merit for various traits based on their genotypic information. Several approaches are used to estimate GEBVs, each with its strengths and limitations. Here are some of the common approaches:

Genomic Best Linear Unbiased Prediction (GBLUP):

·         GBLUP is a popular and widely used approach for estimating GEBVs.

·         It assumes that all marker effects are drawn from a normal distribution with a common variance (known as the genomic relationship matrix).

·         GBLUP estimates marker effects simultaneously using a mixed model framework, incorporating information from all markers to predict breeding values.

·         This approach is computationally efficient and does not require prior knowledge of marker effects or genetic architecture.

Bayesian Methods:

·         Bayesian methods estimate GEBVs by sampling from the posterior distribution of marker effects given the observed data (marker genotypes and phenotypic values).

·         These methods allow for the incorporation of prior information, such as marker effects and population structure, into the analysis.

·         Examples include Bayesian ridge regression, Bayesian LASSO, and Bayesian variable selection methods.

Single-Step Methods:

·         Single-step methods integrate pedigree, genotypic, and phenotypic information into a unified framework for GEBV estimation.

·         These methods use a combined relationship matrix that accounts for both pedigree and genomic relatedness among individuals.

·         Single-step methods offer improved accuracy and robustness compared to traditional methods that rely solely on genomic or pedigree information.

Machine Learning Methods:

·         Machine learning algorithms, such as random forests, support vector machines, and neural networks, can also be used for GEBV prediction.

·         These methods can capture non-linear relationships between markers and phenotypes and handle high-dimensional data effectively.

·         Machine learning approaches may offer advantages for traits with complex genetic architectures or interactions.

Combination of Approaches:

·         Some GEBV estimation methods combine multiple approaches or incorporate additional information, such as functional genomics data or environmental covariates, to improve prediction accuracy.

·         Ensemble methods, which combine predictions from multiple models, can also enhance the robustness and accuracy of GEBV predictions.

In common use, GBLUP is prevalent for GEBV estimation due to several reasons:

·         Robustness: GBLUP is robust to the assumption of the distribution of marker effects and works well under a wide range of genetic architectures.

·         Computational Efficiency: GBLUP is computationally efficient and scalable to large datasets, making it practical for routine genomic selection analyses.

·         Accuracy: Despite its simplicity, GBLUP often provides competitive prediction accuracy compared to more complex methods, especially for traits with polygenic inheritance.

While other methods may offer advantages in certain scenarios, GBLUP remains a popular choice for GEBV estimation in many plant breeding programs due to its balance of accuracy, computational efficiency, and simplicity. However, the choice of approach may vary depending on the specific characteristics of the breeding population, trait of interest, and available computational resources.