Rapid cycling genomic selection in maize landraces

Area of Science:

• Plant genetics and breeding
• Agricultural science
• Genomic selection applications

Background:

• Maize landrace genetic diversity is underutilized in elite germplasm improvement.
• Rapid cycling recurrent pre-breeding schemes are crucial for sustainable crop development.
• Genomic selection offers potential for accelerating pre-breeding in diverse populations.

Purpose of the Study:

• To investigate the efficacy of rapid cycling genomic selection for pre-breeding a maize landrace.
• To assess the impact of genomic selection on early plant development traits.
• To identify factors influencing the success of rapid cycling genomic selection schemes.

Main Methods:

• Trained genomic prediction models on 899 landrace-derived doubled-haploid (DH) lines genotyped with a 600k SNP array across 11 European environments.
• Performed three cycles of genomic selection and recombination for line per se performance in two replications.
• Evaluated 688 DH lines from all cycles and replications for selected and unselected traits in seven environments.

Main Results:

• Achieved significant selection response, increasing traits under directional selection by approximately two standard deviations.
• Observed diminishing realized selection response and declining prediction accuracies across cycles, particularly for directional traits.
• Demonstrated that retraining prediction models improved accuracy in later cycles, and replications showed variations in response and accuracy.

Conclusions:

• Rapid cycling genomic selection is effective for pre-breeding maize landraces, yielding substantial gains in early development traits.
• Prediction accuracy declines across cycles, highlighting the need for model retraining and careful experimental design.
• The study provides key insights for optimizing rapid cycling genomic selection schemes and maximizing their success in crop improvement.