Genetic enhancement of root, tuber and cereal crops via pangenomics, multi-omics integration and AI-driven prediction.

Breeding root, tuber, and cereal crops faces the critical challenge of unlocking extensive genetic variation and addressing complex gene-environment interplays to boost yield, quality, and resilience. Recent technological advances in pangenomics, multi-omics data integration, and artificial intelligence (AI)-driven predictive modeling offer unparalleled opportunities to transform crop improvement. Pangenomics transcends the limitations of single reference genomes by encompassing the full genomic diversity within species, capturing critical structural variations and rare alleles that underpin stress tolerance and productivity traits. When layered with multi-omics datasets spanning genomics, transcriptomics, proteomics, and metabolomics, a holistic insight is gained into molecular networks governing plant adaptation and development. State-of-the-art AI methodologies harness these complex datasets, enabling precise genomic selection, accurate trait prediction, and discovery of novel candidate genes, thereby optimizing breeding pipelines. This review presents current knowledge on how this synergistic approach heralds a new era of climate-smart agriculture, empowering resilient, high-performing cultivars essential for global food security amid escalating environmental uncertainties with a particular focus on root, tuber and cereal crop genetic enhancement through pangenomics and multi-omics integration and AI-driven predictive modeling. Together, these innovations enable tailored breeding strategies that align genetic potential with environmental specificity and farmer needs, while highlighting the remaining hurdles-data standards, model interpretability, computational cost, and equitable access-that must be addressed to realize widespread impact. Demonstrated in staple crops such as maize, rice, wheat, potato, and cassava, this integrated framework accelerates genetic gain by reducing breeding cycles and facilitating allele introgression from wild relatives. The integrative approach also provides a better understanding of resolving persistent hurdles around data standardization, interpretability, computational demands, and equitable technology access. We recommend, (i) training on diverse, field-collected datasets; (ii) integrating envirotyping covariates into genomic selection to quantify G×E interactions; (iii) adopting standardized metadata schemas; and (iv) fostering interdisciplinary collaboration.