Genomics and Bioinformatics for Next-Generation Crop Improvement: From Pangenomes to AI-Driven Precision Breeding

Abstract

Crop genomics has been based on single reference genomes for over last two decades, which have been used for the development of markers, genome-wide association studies and genomic selection, but which have underrepresented structural variation, gene content and haplotype diversity within species.This review explores the impact of the convergence of pangenomics, advanced bioinformatics, artificial intelligence and high dimensional biological datasets on crop improvement.This review follows the evolution of long-read and telomere-to-telomere genome assemblies to pangenomes and super-pangenomes, population and comparative genomics, haplotype based breeding, and genomic prediction for traits and environments.It also assesses the integration of transcriptomic, proteomic, metabolomic, epigenomic, single-cell and spatial omics data with genomic data, as well as high throughput phenotyping and enviromics.A special focus is on machine learning, deep learning, explainable AI, foundation models and pangenome-guided genome editing for gene discovery, phenotype prediction, sequence design and de novo domestication.Applications for yield stability, abiotic stress tolerance, nutrient use efficiency, disease resistance, and biofortification are critically evaluated.The persistent challenges of computational bottlenecks, unequal genomic resources and the disconnect between genomic discoveries and breeder ready tools are also explored.This review focuses on the shift towards an integrated, prediction-based approach to the design of optimal genotype-phenotype combinations under various production environments.