Genomic prediction in early selection stages using multi-year data in a hybrid rye breeding program.

Area of Science:

• Plant breeding
• Quantitative genetics
• Genomic prediction

Background:

• Multi-year data offers valuable marker effect information for genomic prediction (GP).
• Single-year GP models are simpler but may miss crucial multi-year genetic interactions.
• Replication at marker loci across years allows for sophisticated multi-year analysis strategies.

Purpose of the Study:

• To evaluate different GP approaches for modeling genotype-by-year (GY) effects and breeding values simultaneously using multi-year data.
• To assess the impact of various scenarios, including training/validation set relatedness and genotype selection, on predictive ability.
• To present prediction approaches emphasizing kinship for modeling GY effects.

Main Methods:

• Utilized empirical grain yield data from a rye hybrid breeding program.
• Compared GP models with and without kinship for modeling genotype-by-year (GY) effects.
• Evaluated models under scenarios with varying degrees of relatedness between training and validation sets and selected training populations.

Main Results:

• Models incorporating kinship to account for genotype-by-year (GY) effects showed an average 5% increase in predictive ability, especially for disconnected datasets.
• Utilizing data from multiple selection stages significantly increased predictive ability by approximately 30% when predicted genotypes were closely related to the training set.
• Selecting top-yielding genotypes combined with kinship for GY modeling improved predictive ability in multi-cycle datasets.

Conclusions:

• Multi-year data and appropriate modeling are essential for GP, enabling the separation of GY effects from genomic estimated breeding values.
• Model selection and ensuring sufficient relatedness between predicted candidates and training set genotypes are critical for successful GP.
• The study highlights the benefits of advanced modeling strategies for optimizing genomic prediction in plant breeding programs.