Back to search

University of Minnesota

Advancing predictive breeding in maize: Genomewide prediction of stability across environments and artificial intelligence-based emulation of breeder decisions

Abstract

dc:description.abstract

Predictive breeding in maize (Zea mays L.) currently focuses on trait means. This dissertation examined two extensions of predictive breeding. The first study evaluated whether genomewide prediction could accurately predict hybrid stability across environments. In seven maize testcross populations evaluated in 10--18 environments, Eberhart--Russell stability slopes varied substantially. However, predictive ability for stability was low and inconsistent, ranging from -0.07 to 0.41 for grain yield, -0.08 to 0.52 for moisture, and -0.18 to 0.62 for test weight. Predictive ability was always lower for stability than for across-environment means. Multiple genomewide prediction models produced similar results, and low split-half correlations indicated that a low signal-to-noise ratio largely constrained predictability. Predicting stability was therefore too inconsistent to be routinely useful. The second study assessed whether machine learning could emulate selection decisions based on multi-trait phenotypic summaries. Seven experienced maize breeders independently selected candidates within the same seven populations. Selection decisions were modeled using two linear and four nonlinear machine learning algorithms. In across-population prediction, the probability that a random selected candidate was ranked ahead of a random non-selected candidate approached 0.90 for the best-performing models, and about 70--80% of breeder decisions were exactly replicated. Gradient boosting and logistic regression performed best, whereas random forest and k-nearest neighbors performed poorest. Selection was strongest for grain yield and was weaker for moisture, test weight, plant height, and ear height. The results showed that artificial intelligence via machine learning can effectively mimic selection decisions in maize and be used as a decision-support system for breeders.

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Ariza Hernandez, Isaias

Subjects

dc:subject × 4

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Citation
Ariza Hernandez, Isaias. (2026). Advancing predictive breeding in maize: Genomewide prediction of stability across environments and artificial intelligence-based emulation of breeder decisions. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/280355.

Chain of custody

source
Harvested from
University of Minnesota
Base URL
conservancy.umn.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Ariza Hernandez, Isaias. Advancing predictive breeding in maize: Genomewide prediction of stability across environments and artificial intelligence-based emulation of breeder decisions. 2026. https://hdl.handle.net/11299/280355