Sparse multi-trait genomic prediction under balanced incomplete block design.

Area of Science:

• Agricultural Science
• Genetics
• Plant Breeding

Background:

• Genomic selection (GS) is a powerful tool in plant breeding for predicting genetic merit.
• Sparse testing, evaluating fewer genotypes across environments, enhances GS efficiency and reduces costs.
• Optimal allocation of lines to environments is crucial for maximizing prediction accuracy in sparse testing scenarios.

Purpose of the Study:

• To evaluate four distinct methods for allocating lines to environments under sparse testing.
• To compare the prediction power of these allocation methods in a multi-trait genomic prediction context.
• To identify the most effective strategy for optimizing resource allocation in plant breeding programs.

Main Methods:

• Four allocation methods were assessed: M1 (fraction of lines in all locations), M2 (fraction of lines with shared lines, non-BIBD), M3 (random subset allocation), and M4 (subset allocation using BIBD principle).
• The methods were evaluated using seven real-world multi-environment trial datasets.
• Performance was measured by prediction power within a multi-trait genomic prediction framework.

Main Results:

• Method M4, utilizing the balanced incomplete block design (BIBD) principle, demonstrated the highest prediction power.
• Method M1, allocating a fraction of lines across all locations, performed the worst.
• Methods M3 (random allocation) and M4 showed comparable and superior performance to M1 and M2.

Conclusions:

• Sparse testing strategies employing the BIBD principle (M4) or random allocation (M3) are highly efficient for multi-trait genomic prediction.
• These methods offer a practical approach to increase the efficiency of genomic selection in plant breeding.
• The findings support the adoption of BIBD or random allocation for optimizing genotype evaluation in multi-environment trials.