Lettuce is one of the most valuable, widely consumed vegetable crops and a major contributor to the quality of the american diet

The Department of Agriculture National Institute of Food and Agriculture awarded the University of California, Davis $4.968 million on July 1, 2026, under the Specialty Crop Research Initiative (CFDA 10.309) to develop and validate artificial intelligence foundation models for lettuce crop improvement and to advance genome editing and crop management technologies across California, Arizona, and Florida production regions through June 30, 2031.

The project encompasses four work packages. The first prepares genomic, developmental, physiological, and phenotypic datasets from cultivated lettuce and wild relatives, then trains an AI foundation model to enable breeders to predict phenotypic consequences of allelic variation, guide superior genotype selection, and forecast outcomes of gene editing. The second work package uses genome editing to knock out domestication genes in wild accessions with demonstrated abiotic stress tolerance, generating germplasm with cultivated phenotypes suitable for breeding programs. The third package leverages AI to enhance accessibility and adoption of CropManage software, a tool designed to improve water and nitrogen use efficiencies in lettuce production. The fourth investigates the economics of adopting both CropManage and genome-edited lettuce varieties. Work is performed at UC Davis, Davis, California, with results aimed at increasing production efficiency and contributing to regional agricultural economies.

This highly collaborative, multi-state, long-term project provides the foundation of public and private sector lettuce improvement programs. IT is currently challenging to exploit the vast amounts of genomic, developmental, physiological, and phenotypic data for cultivated lettuce and its wild relatives that we and others have generated. Therefore, we will train an AI foundation model on these diverse datasets in order be able to use these data resources for lettuce improvement. There will be four work packages. The first will focus on preparing the data and training the foundation model, followed by experimental validation of the model. This will greatly assist breeders to work in this data-rich environment. IT will assist in predicting the phenotypic consequences of allelic variation as well as help in the selection of superior genotypes; IT will also predict the consequences of gene editing. The second work package will use genome editing to domesticate wild accessions that we have previously shown to exhibit tolerance to abiotic stressors by knocking out known domestication genes. This will provide germplasm with abiotic stress tolerance that has a cultivated phenotype and therefore can be more readily used in breeding programs. The third work package will use AI to make the successful cropmanage software more accessible and increase its adoption. This will immediately impact water and nitrogen use efficiencies of lettuce crop production. The fourth work package will investigate the economics of adoption of crop manage and of genome edited lettuce. This project will result in more efficient production of high-quality lettuce for the consumer as well as making major contributions to the economies of lettuce producing areas in california, arizona, and florida.