Scalable Sparse Testing Genomic Selection Strategy for Early Yield Testing Stage.

Area of Science:

• Plant breeding
• Quantitative genetics
• Genomic selection

Background:

• Genomic selection (GS) is crucial for accelerating crop improvement.
• Scalable GS strategies are needed for preliminary yield trials.
• Incorporating genotype by environment interaction (GEI) is key for accurate genomic prediction.

Purpose of the Study:

• To explore optimal methods for incorporating GEI in genomic prediction models for scalable GS.
• To evaluate two cross-validation schemes (CV1 and CV2) for predicting genetic merit in maize breeding.
• To assess the impact of environment grouping and historical data on prediction accuracy.

Main Methods:

• Two cross-validation schemes (CV1 and CV2) were evaluated.
• CV2 used the coefficient of determination (CDmean) to optimize subsets of full-sib families per environment.
• Environments were grouped by similar growing/management conditions.

Main Results:

• CV1 and CV2 showed similar prediction accuracies.
• CV2 offers better representation across environments and facilitates robust historical data building.
• Grouping environments improved prediction accuracy and reduced computational load.
• Complementing calibration sets with optimized historical data enhanced prediction accuracy.

Conclusions:

• A scalable, parsimonious approach for multi-environmental trials and early-stage GS was developed.
• CV2 is advantageous for efficient information use across environments and building historical data.
• Environment grouping and optimized historical data are effective strategies for improving genomic prediction accuracy in maize breeding.