The importance of phenotypic data analysis for genomic prediction - a case study comparing different spatial models...

Area of Science:

• Agricultural Science
• Plant Breeding
• Genetics

Background:

• Genomic prediction is vital for plant breeders, relying on genotypic data for selection decisions.
• Accurate genomic prediction requires sufficient genotype data, necessitating combining data across years.
• Phenotypic data analysis is crucial for calibrating genomic prediction procedures.

Purpose of the Study:

• Assess the benefit of spatial models for genomic prediction accuracy.
• Identify optimal stage-wise analysis approaches for weakly connected multi-environment trials (MET).
• Explore genomic prediction for selecting phenotypic data analysis models.

Main Methods:

• Compared stage-wise analysis approaches for a rye MET dataset.
• Utilized various spatial models to improve estimates and predictive abilities.
• Analyzed a weakly connected multi-environment trial (MET) dataset across years.

Main Results:

• Complex spatial models did not significantly enhance genomic prediction accuracy.
• Incorporating row and column effects resulted in the highest predictive abilities.
• Analyzing each year separately with year as a fixed effect provided realistic predictive abilities for poorly connected METs.

Conclusions:

• Combining data across years using year means is effective for weakly linked datasets.
• Fitting row and column effects captures significant field trial heterogeneity.
• Predictive abilities can guide the selection of phenotypic data analysis models, aligning with but distinct from traditional criteria like AIC.