Discover the SciOpen Platform and Achieve Your Research Goals with Ease.
Search articles, authors, keywords, DOI, etc.
Multi-environment trials (METs) are essential for maize breeding, but their implementation is often constrained by limited resources, particularly the high cost of phenotyping. To optimize resource allocation, 264 testcrosses derived from 100 inbred lines and three testers were evaluated across six locations. Four genomic prediction models, five cross-validation schemes (CV1–CV5), and five training set proportions (S1–S5) were systematically assessed for predicting grain moisture content (GMC), grain yield (GY), and plant height (PH). Together, the five cross-validation schemes and five training set proportions generated 25 sparse testing strategies. Among the four models, the reaction norm model consistently achieved the highest prediction accuracy, and was therefore selected for subsequent analyses. Averaged across S1–S5, CV1 showed the lowest prediction accuracy, with mean accuracies of 0.361 for GMC, 0.413 for GY, and 0.480 for PH. In contrast, CV2–CV5 achieved higher and comparable accuracies. Increasing the training set proportion from 1/6 (S1) to 5/6 (S5) improved prediction accuracy by 48.3% for GMC, 30.3% for GY, and 26.7% for PH, although further gains became limited once the training set proportion exceeded approximately 50% (S3). In addition, CV3 enabled reliable estimation of parental general combining ability based on predicted hybrid performance, particularly at higher training set proportions. Overall, integrating multi-environment sparse testing with genomic prediction enables efficient hybrid evaluation under limited resources, providing a cost-effective strategy to reduce phenotyping demands and accelerate breeding progress.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Comments on this article