AI-Driven Genomic Selection and Precision Breeding for Accelerating Climate-Resilient Crop Improvement: A Multi-Environment Genome-Wide Prediction Study

Abstract

Climate instability threatens crop productivity by exposing breeding populations to drought, heat, and increasingly variable growing conditions. This study evaluated an artificial intelligence-driven genomic selection framework for accelerating the identification of climate-resilient crop genotypes across diverse environments. A simulated multi-environment panel comprising 1,240 breeding lines was evaluated across twelve contrasting field environments, generating 14,880 line-by-environment observations for grain yield, drought tolerance, canopy temperature, flowering time, and biomass. Genome-wide information from 1,000 single-nucleotide polymorphism markers was integrated with climate and soil covariates using genomic best linear unbiased prediction, random forest, gradient boosting, and deep neural network models. Prediction performance was assessed through repeated five-fold cross-validation, leave-one-environment-out validation, and independent testing. The deep neural network consistently achieved the highest independent-test accuracy, reaching 0.747 for grain yield, 0.732 for drought tolerance, 0.668 for canopy temperature, 0.793 for flowering time, and 0.712 for biomass. Incorporating environmental covariates improved prediction accuracy across all models and validation procedures, with gains ranging from 0.052 to 0.060 correlation units. Twenty stable genomic loci were identified across the ten chromosomes, with stability scores ranging from 0.746 to 0.940 and consistent effects across eight to twelve environments. These loci were associated with stress signalling, root development, photosynthetic efficiency, and yield maintenance. Overall, the findings demonstrate that integrating artificial intelligence, genomic markers, and environmental information can improve predictive accuracy, support early selection of superior genotypes, and potentially reduce the time and cost required for climate-resilient crop improvement.