抄録
The increasing frequency of drought, salinity, and temperature extremes under global climate change poses a major constraint on sweet potato productivity and stability, primarily through complex, multi-scale disruptions in stress-response regulatory networks. Improving stress resilience and environmental adaptation in sweet potato, therefore, requires more precise identification of functional genetic determinants and more efficient integration of multi-source biological and environmental data. In this context, artificial intelligence (AI) offers new opportunities to address limitations of conventional breeding approaches, particularly in disentangling complex genotype–phenotype–environment interactions. This review synthesizes recent advances in AI-enabled stress-resilience improvement, focusing on deep learning-based stress-resistance gene discovery, optimization of adaptive trait prediction, and AI-guided selection of gene-editing targets. It further highlights integrative frameworks that couple machine learning models with gene-editing technologies to enhance the accuracy of target identification under multi-stress conditions. In addition, emerging AI-driven breeding strategies that simulate plant performance across heterogeneous environmental scenarios are discussed, emphasizing their potential to accelerate data-informed cultivar development. Key challenges remain in model generalizability, biological interpretability, and multi-environment validation. This study provides a consolidated perspective on the role of AI in decoding stress-response complexity and supports the development of next-generation intelligent breeding systems for climate-resilient sweet potato improvement.
