Genomic selection using random regression to predict hybrid performance in untested environments by leveraging environmental data

Abstract

Genotype‐by‐environment interaction (G×E) remains one of the major challenges to achieving genetic gains in tropical maize ( Zea mays L.) breeding. To address this, we implemented the geographic information system–factor analytic (GIS‐FA) framework, which integrates factor‐analytic (FA) models with environmental covariates and genomic kernels to predict hybrid performance in untested environments. Unlike traditional G×E models, GIS‐FA links environmental descriptors directly to the latent factor loadings estimated by FA models. These loadings are then extrapolated to new environments using partial least squares regression, enabling accurate predictions of genotypic performance across untested locations. By incorporating a nonlinear Gaussian genomic kernel into the FA framework, potentially capturing additive and nonadditive genetic effects, predictive accuracy improved by up to 13.0% relative to the environment‐only model across four cross‐validation schemes. This integrative approach enabled the development of high‐resolution predictive maps that translate complex G×E patterns into visual tools, supporting decision‐making in hybrid positioning. The results highlight the practical advantages of GIS‐FA for tropical maize breeding programs: (1) robust prediction of hybrid performance beyond experimental networks, (2) strategic genotype recommendation in the target population of environments, and (3) improved accuracy of site‐specific predictions through the integration of genomic information. In summary, GIS‐FA provides a scalable, data‐driven framework to guide hybrid recommendation, deployment, and resource allocation in tropical maize breeding.