Predicting the accuracy of genomic predictions

Area of Science:

• Animal Breeding and Genetics
• Quantitative Genetics
• Genomic Selection

Background:

• Genomic prediction requires accurate mathematical models for breeding program design.
• Deterministic models for pedigree-based estimates of breeding values (PEBV) exist, but lack accuracy prediction for genomic selection.
• Predicting the accuracy of genomic estimated breeding values (GEBV) in selection candidates is a critical missing component.

Purpose of the Study:

• To develop a deterministic method for predicting GEBV accuracy in selection candidates within a closed breeding population.
• To establish a method based on reference population accuracy and candidate-ancestor distance.
• To provide a tool for optimizing genomic selection strategies.

Main Methods:

• Modeled GEBV accuracy as a combination of PEBV accuracy and genomic relationship-based EBV (DEBV) accuracy.
• Quantified the loss of DEBV accuracy using the effective number of independent chromosome segments (M_e).
• Compared M_e estimation methods (Fisher information, selection index, variance of relationships) via simulation.

Main Results:

• Both Fisher and selection index approaches accurately predicted target population accuracy over time.
• The selection index approach yielded M_e estimates less influenced by heritability, reference size, and selection, making it a more robust population parameter.
• Variance of relationships underpredicted M_e and was highly sensitive to selection.

Conclusions:

• A deterministic method to predict GEBV accuracy in closed populations was successfully developed.
• The population parameter M_e can be derived from reference data and applied across datasets and traits.
• This method aids in evaluating genomic prediction benefits and optimizing breeding programs.