Predictive Crop Performance helping make NDSU breeding programs more efficient

Food sustainability is a crucial component of the North Dakota State University Food, Energy and Water Security initiative, and Predictive Crop Performance is a way to ensure those public sector crops are grown now and in the future.

Predictive Crop Performance (PCP) is defined as the use of data, modeling and technology to forecast how a crop will grow, develop and yield in a specific environment. The goal of this and other initiatives, such as the Big Data Initiative at NDSU, is to collect data, set up experiments, and develop PCP tools to ensure that public crops are sustainable against current and future abiotic and biotic stresses.

“The long-term goal of this project is to utilize Predictive Crop Performance tools to ensure North Dakota's public sector crops are resilient,” said Richard Horsley, NDSU plant sciences head and barley breeder.

Public sector crops are those where private industry does not invest significantly in breeding efforts. At NDSU, these include barley, dry beans, flax, oats, and pulse crops, Horsley said.

The United States Department of Agriculture (USDA) funding supports the sequencing and data gathering for the genomes of the crops within the project. The National Science Foundation Engine Type-2 funding supports the acceleration of data tools and technologies to aid research and development in data management and data analytics workflows, which in turn support the development of resilient food cropping systems.

“This Predictive Crop Performance project shows how powerful data and analytics can be in advancing agriculture,” said North Dakota Agricultural Experiment Station associate director Frank Casey. “It’s helping researchers use digital tools to make faster, better decisions. Our Agricultural Data Analytics group is driving this progress and setting an example for the land-grant system.”

The NDSU Agricultural Data Analytics Team developed the predictive tools being used in the program. The initial crops being worked with are barley, dry beans, dry peas and oats.

“To achieve this, genotypic and phenotypic performance data will be used to predict which early generation breeding lines should be advanced to replicated yield trials across multiple environments,” Horsley said. “Additionally, genomic and phenomic data will be used to improve selection decisions and determine which crosses are predicted to produce lines with the desired agronomic traits, disease resistance, and end-use quality sought by farmers and end-users.”

The primary objective of this PCP project is to develop tools that can accurately predict crop growth in North Dakota, even as environmental conditions change. These tools will help ensure that publicly developed crops in the state remain strong and productive, despite challenges such as pests, diseases, drought, or heat.

"Working the barley plots showed me how field data and analytics can directly help breeders choose lines that perform under real-world scenarios,” said Mat Souza, who is an NDSU graduate student and a UAV software engineer for the NDSU Data Analytics Team. “It’s exciting to see our images and data turn into decisions for the next season.”

To achieve this, the project will utilize data on the growth characteristics (traits) and genetic makeup of various crop varieties to predict which ones are most likely to perform well under future climate conditions. This will include studying the full range of genetic differences in public crops grown in North Dakota by building detailed maps of their DNA (genomes) using samples from NDSU breeding programs; understand how different breeding lines in the NDSU programs vary from each other and finding the specific genes that help crops survive and thrive in changing environments, such as extreme weather or new pests and diseases.

“Genomic selection, a form of predictive crop performance, improves the efficiency of our breeding programs,” Horsley said. “While this method does not shorten the time needed to develop a new variety, it increases the number of lines available at the end of the breeding process that are candidates for release.”

In 2023, Horsley designed a 3.5K AgriSeq SNP array to use in genomic selection by the NDSU barley breeding program.

“Using the genotype data and historical phenotype data, I selected about 500 lines each year to advance to replicated, multi-location yield trials in North Dakota based on predicted values for 10 traits, including those related to agronomic performance and end-use quality,” Horsley said. “I also use the genotype data to decide which crosses I will make each fall.”

A crucial characteristic in malt quality is the concentration of wort β-glucan. High levels of this soluble fiber contribute to the cholesterol-lowering effects of oats in humans; however, in barley malt, high wort β-glucan levels are undesirable because they slow filtration during packaging and brewing. From 2017 to 2024, the NDSU barley breeding program has utilized genomic selection to reduce wort β-glucan levels, resulting in an increased number of lines advancing to the preliminary yield trial with lower β-glucan levels. That includes several with lower wort B-glucan levels than the breed ND Genesis.

“Before using genomic selection, we tested 1,200 lines annually in our preliminary yield trials, with about 400 of those lines (33%) having less wort β-glucan than ND Genesis,” Horsley said. “Currently, we test around 500 lines each year, and approximately 440 (88%) of them show wort β-glucan levels lower than ND Genesis. Genomic selection has significantly reduced breeding costs by decreasing the number of plots needed.”

In 2024, Phil McClean and Juan Osorno designed a 3.9K AgriSeq SNP array to predict crop performance in dry beans through genomic selection.

“They are using the SNP array to genotype early-generation dry bean lines to determine which ones to advance to replicated, multi-location yield trials,” Horsley said. “Traits predicted using the genomic data include canning quality, resistance to common bacterial blight, days to flowering, yield, days to maturity, plant height, and resistance to white mold.”

Genomic selection plays an essential role in predicting crop performance and has made NDSU’s breeding programs more efficient, Horsley said.

“While this method does not shorten the time needed to develop a new variety, it increases the number of lines available at the end of the breeding process that are candidates for release,” Horsley said. “The NDSU Agricultural Data Analytics Team is developing tools that enable our NDSU breeding programs to employ the same breeding methods used by large multinational companies in the creation of new corn hybrids and soybean varieties. Therefore, our breeding programs will become more efficient in producing the resilient crop varieties that North Dakota farmers need.”