Location: Cereal Crops Research
Project Number: 3060-21000-038-051-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Jul 1, 2022
End Date: Jun 30, 2023
Spring wheat is vulnerable to Fusarium head blight (FHB) or scab, and the most economical way of combating the disease is to grow varieties with genetic resistance. Genomic selection (GS) is a promising molecular breeding method to predict performance and accelerate the development of FHB resistant spring wheat cultivars that also meet performance requirements. A challenge with this method is that predicted trait metrics need to be generated rapidly every year to inform selection/crossing decisions in the next generation. The overall goal of the project is to provide centralized resources to enable GS in all the northern US spring wheat breeding programs. Towards this goal, a specific objective pursued in this project is the integration of phenotypes and genotypes from different programs and the development of initial GS models to support rapid performance prediction and selection decisions. Currently, the four largest public spring wheat breeding programs (University of Minnesota, North Dakota State University, Montana State University, and South Dakota State University) operate independently to evaluate breeding germplasm traits—such as yield, end-use quality, and disease, including FHB. Genomic selection (GS) is in various stages of usage for each program, and the intent of this project is to increase collaboration among the programs with greater shared trait measurements and shared resource allocation for GS modeling. If successful, this project will not replace, but augment and improve current GS activities in the supported programs.
To initiate the project, we will utilize the existing infrastructure of the USDA-ARS Uniform Regional Nursery (URN), which is a useful 14 location trial for shared evaluation of public and private lines that has been running since 1960. The spring wheat breeders manage the URN locations in their regions, and each location evaluates 10-20 public entries per year. This data, along with per-program data will be combined and jointly analyzed to identify trait-associated markers. Additionally, the initial goal of the project is to build a prediction model in the programs that don’t already have it and improve it in the ones that do. Since it will take several years of data (and potentially more lines) to be able to utilize the expanded URN alone for prediction, we will investigate using prior years’ data from each program as well. In many cases, individual breeder trials are co-located with URN trials, and marker effects will likely be relevant across these projects for future incorporation. We will obtain records from the previous four years of trials that had performance and FHB disease ratings measured. Each program has differently sized trials that contain this data, and we expect to build up and refine training models that contain between 300-800 lines. Remnant seed will be genotyped with the new low-density array if necessary, or imputation will be used to harmonize disparate genotyping platforms together using the extensive amount of sequencing our group is currently performing. Initial genomic selection models will be developed and evaluated with cross- and forward-validation. With centralized data management, many iterations of analysis can be performed, like the prediction of URN trait data with a training model that only contains lines from a single program or vice versa. Relative to a given program, the URN trial data will provide greater genetic diversity—which will potentially improve model accuracy. Additionally, the large number of locations will also allow us to test region-specific accuracy with higher resolution. The USDA-ARS also manages a Uniform Regional Scab Nursery that operates much like the URN but emphasizes FHB resistance, especially among new sources. Lines in this nursery often have one or more traits that make them unacceptable as a candidate for cultivar release. We will also explore incorporating this germplasm into our modeling to: (1) predict crosses that better incorporate these resistance sources into adapted breeding material, and (2) use the genotyping data of URSN entries to identify marker effects that predict FHB resistance better. At the end of the first year, the goal is to obtain enough information from previous years so the prediction can benefit all four programs. We will then begin with the following timeline to fulfill a “one stop shop” model of centralized services.