MACHINE LEARNING-ASSISTED PREDICTION OF DROUGHT RESISTANCE TRAITS IN CROP GERMPLASM USING GENOMIC AND PHENOTYPIC DATA

Abstract

Efficient identification of drought-resistant crop varieties remains a major challenge in conventional breeding programs due to complex genotype–environment interactions. This study developed a machine learning-based prediction framework integrating genomic and phenotypic datasets to improve drought resistance selection in crop germplasm. A diverse germplasm collection was evaluated under drought stress conditions, and multiple morphological, physiological, and yield-related traits were combined with genome-wide molecular marker information. Several machine learning algorithms, including random forest, support vector regression, and gradient boosting models, were trained and compared for predicting drought resistance performance. The optimized model demonstrated high prediction accuracy and successfully identified key phenotypic and genotypic features contributing to drought adaptation. Feature importance analysis revealed major predictive traits associated with water-use efficiency, root development, antioxidant activity, and stress-responsive genetic variation. The integration of artificial intelligence with genomic selection enhanced prediction capability compared with traditional phenotype-based approaches. This study demonstrates the potential of machine learning-assisted breeding to accelerate the identification of drought-resistant germplasm and reduce breeding cycles. The proposed framework provides a scalable strategy for integrating computational approaches into climate-smart crop improvement programs.