Harnessing big data for enhanced genome-wide prediction in winter wheat breeding

Area of Science:

• Plant breeding
• Genomics
• Agricultural science

Background:

• Genomic selection (GS) is crucial for accelerating crop improvement.
• Training population size and diversity are key factors influencing GS accuracy.
• Winter wheat breeding programs often have fragmented data.

Purpose of the Study:

• To assess the impact of combining diverse datasets on genomic prediction accuracy in winter wheat.
• To evaluate the potential of big data for predictive breeding.
• To determine if a larger, more diverse training population improves prediction of grain yield and plant height.

Main Methods:

• Assembled a large-scale winter wheat dataset (approx. 18,000 inbred lines, 250,000 plots).
• Trained genome-wide prediction models using combined public and private breeding data.
• Compared prediction accuracy against models trained on individual datasets.
• Utilized data from post-registration trials across diverse environments.

Main Results:

• Combined big data significantly enhanced prediction accuracy: up to 97% for grain yield and 44% for plant height.
• Prediction ability improved substantially compared to using individual training sets.
• The expansion of training set size and genetic diversity were primary drivers of improvement.
• Genome-wide prediction accuracy increased with larger training population size.

Conclusions:

• Big data integration from multiple breeding programs enhances genomic prediction in winter wheat.
• This approach significantly boosts predictive ability for key agronomic traits.
• Big data is a powerful tool for accelerating genetic gain in predictive breeding programs.
• The findings support the adoption of big data strategies for efficient winter wheat improvement.