Genome-based prediction of testcross values in maize

Area of Science:

• Plant breeding
• Genomics
• Quantitative genetics

Background:

• Advanced cycle breeding populations require accurate prediction of genetic values for untested lines.
• Genome-based prediction offers a promising alternative to traditional pedigree-based methods.
• Evaluating prediction models in large-scale experimental settings is crucial for breeding programs.

Purpose of the Study:

• To assess the efficacy of genome-based prediction models for testcross values in an advanced maize breeding population.
• To compare the predictive performance of genomic models against pedigree-based models.
• To investigate the impact of relatedness and sample size on prediction accuracies.

Main Methods:

• Utilized best linear unbiased prediction (BLUP) and stratified cross-validation.
• Genotyped 1,380 doubled haploid maize lines using 1,152 single nucleotide polymorphism (SNP) markers.
• Phenotyped for grain yield and grain dry matter content across seven locations.

Main Results:

• Genomic models significantly improved prediction accuracies compared to pedigree-based models.
• High prediction accuracies (0.72-0.74) were achieved for grain yield with genomic data.
• Prediction accuracies decreased with increasing genetic distance between estimation and test sets (0.47-0.48).

Conclusions:

• Genome-based prediction is effective for estimating the genetic value of untested maize lines in advanced breeding.
• Genomic selection can be successfully implemented in maize breeding programs.
• Prediction accuracy is influenced by the relatedness between individuals in the training and testing sets.