A new study led by the International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), in collaboration with the Leibniz Institute of Plant Genetics and Crop Plant Research (IPK), Germany, and the University of Queensland, Australia, has proposed a new framework that could reshape crop breeding by integrating biology-driven pangenetics with artificial intelligence (AI)-powered prediction models.
Published in Molecular Plant, the study outlines a forward-looking strategy for harnessing rapidly evolving AI prediction models to overcome the complexities of breeding the next generation of crops.
The researchers argue that combining functional genomics, genebank diversity and advanced AI could accelerate the development of climate-resilient, high-yielding, nutritious and market-ready crop varieties capable of meeting future agricultural challenges.
The authors note that modern crop breeding has advanced through two complementary but largely separate approaches. Molecular genetics focuses on understanding the biological mechanisms that control plant traits, while quantitative genetics relies on statistical prediction using large-scale genomic and phenotypic data.
Although both have contributed significantly to crop improvement, the study suggests that integrating the two approaches offers new opportunities to design crop varieties tailored to specific environments, production systems and market needs.
The study also highlights a long-standing dilemma in breeding research. While genomic prediction effectively captures statistical patterns associated with complex traits such as grain yield, it often overlooks the biological relationships among component traits, including interactions and trade-offs between yield, grain quality and stress tolerance.
Functional biology-based approaches, on the other hand, provide mechanistic insights into trait regulation but are typically validated in limited genetic backgrounds and environmental conditions.
To bridge this gap, the researchers advocate the concept of pangenetics, an emerging framework that extends beyond pangenomics by linking genomic variation with functional trait biology. These biologically informed genomic features can then be incorporated into AI-driven prediction models, including deep learning frameworks, to improve prediction accuracy, identify superior breeding lines more efficiently and accelerate genetic gain.
“This strategy presents a transformative vision for crop breeding. When AI-powered prediction is integrated with functional genomics and the vast diversity conserved in genebanks, it can shape the next era of crop improvement. At ICRISAT, we are advancing AI-enabled approaches across the entire agri-food value chain to develop future-ready technologies and services for dryland agriculture, contributing to global food security,” said Dr Himanshu Pathak, Director General of ICRISAT.
Dr Stanford Blade, Deputy Director General – Research & Innovation at ICRISAT, said the institute’s breeding programmes already generate extensive genomic and phenomic datasets that could benefit from the proposed approach.
“ICRISAT’s breeding programs span six crops and generate vast amounts of genomic and phenomic data. By integrating AI with biology, as outlined in this roadmap, we can translate biological discoveries into scalable crop improvement solutions in a much shorter time frame,” Blade said.
Dr Raman Babu, Global Research Program Director – Accelerated Crop Improvement at ICRISAT, said the institute was already pursuing innovations to improve breeding efficiency.
“ICRISAT continues to refine its breeding programs to make them more efficient. Our recently developed Speed Breeding Protocols are one example. This represents another promising direction through which we can deliver superior crop varieties to farmers with greater speed, precision, and efficiency as these concepts move from theory to practical implementation,” he said.
The study says realizing the proposed framework will require stronger interdisciplinary collaboration across genomics, breeding, physiology, AI and data science, as well as investments in data-sharing platforms, responsible governance of genomic resources and capacity building for future researchers.
Dr Manish K. Pandey, Principal Scientist – Genebank & Trait Discovery at ICRISAT, said AI should be viewed as more than a computational tool.
“The study positions AI not merely as a computational tool, but as a unifying framework capable of connecting biological discovery with scalable breeding decisions to accelerate crop improvement for a climate-challenged world,” Pandey said.
“We are already advancing key components of this vision through large-scale genebank genomics, pangenome development, trait discovery, and AI-assisted predictive breeding research.”
The work underpinning the strategy paper was supported by the Department of Biotechnology, Government of India; the Indian Council of Agricultural Research through the ICAR–ICRISAT collaborative programme; the Gates Foundation; the Crop Trust; the CGIAR Genebanks Accelerator; the CGIAR Science Program on Breeding for Tomorrow; and the global initiative Vision for Adapted Crops and Soils (VACS).
