Enhancing genomic prediction with genome-wide association studies in multiparental maize populations

Area of Science:

• Plant genetics
• Quantitative genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify nucleotide variants affecting complex traits, but many have small effects and low repeatability.
• Genomic prediction (GP) models typically use all markers, unlike GWAS which focuses on individual loci.
• Integrating GWAS results into GP models is crucial for improving prediction accuracy, especially in complex traits.

Purpose of the Study:

• To develop and evaluate an integrative modeling approach combining GWAS and GP for complex traits in maize.
• To compare different association test models and their impact on GP accuracy.
• To assess the effectiveness of incorporating significant single-nucleotide polymorphisms (SNPs) as fixed effects in GP models.

Main Methods:

• Compared association tests with SNP effects constrained vs. varying across families.
• Integrated significant association SNPs as fixed effects into a GP model with random polygenic background effects.
• Utilized simulation studies and cross-validation on maize nested association mapping population data.

Main Results:

• The effectiveness of the integrative approach depends on trait polygenicity.
• GP models incorporating significant SNPs and polygenic background improved prediction accuracy for moderately complex traits.
• Highly polygenic traits did not show significant improvement in GP accuracy with this method.

Conclusions:

• Individual SNPs with strong association signals can effectively enhance genomic prediction.
• The proposed integrative modeling approach offers a novel strategy for reliable gene discovery and robust genomic prediction in maize.
• This method holds promise for improving marker-assisted selection and breeding strategies.