High school yearbooks often feature a section dedicated to senior superlatives, spotlighting graduating students whose peers have voted them most likely to succeed or become famous — or maybe even most likely to be late to class. With some help from technology, LSU AgCenter researchers are now able to do the same for prospective rice varieties.

Using a plant breeding approach called genomic selection, scientists can use DNA information to predict which experimental rice lines have the best odds of offering desirable traits like strong yields, resistance to diseases like sheath blight and high milling quality. They also can weed out lines that are most likely to have problems.

“We use DNA markers to understand the relationships between all the lines, and, based on those relationships, we can determine if the line is likely to be good for the trait or bad for the trait,” explained Adam Famoso, a rice breeder and director of the AgCenter H. Rouse Caffey Rice Research Station. “We’ve explored this and have been optimizing it for years. Now we’ve taken the next step. We’re fully integrating it into the breeding program.”

This advance will save time and resources that in the past would have been spent on growing test plots to evaluate the performance of newly crossed rice plants.

Genomic selection is done by taking small tissue samples from the leaves of rice plants. These samples are managed by Brijesh Angira, an associate professor in the rice station’s breeding marker lab, and sent to a third-party provider for analysis using the AgCenterdeveloped LSU500 DNA marker set. These data points are used to characterize the relationships of the lines for use in genomic predictions.

It isn’t all that difficult to generate data about the DNA of the plant samples. “What’s challenging is making sense of that data,” Famoso said. He praised the efforts of Roberto Fritsche Neto, an assistant professor at the rice station who played a key role in conducting the statistical modeling needed to craft the marker set, and postdoctoral researcher Jomar Punzalan and doctoral student Kashish Grover.

“Each of them has conducted extensive optimization studies to determine the best datasets for genomic selection,” Famoso said. “Now it’s routine. Those statistical pipelines are already set up for us to use.”

Genomic selection will complement marker-assisted selection, another high-tech method for screening potential rice varieties that’s already in use at the rice station. The difference between the two? Markerassisted selection focuses on specific individual genes while genomic selection is useful for traits that are controlled by more than one gene.

In this study, 250 lines were evaluated for yield in 2020. The top and bottom 20% were selected based on 2020 field performance. In addition, the entire population was predicted using two genomic selection datasets. The top and bottom 20% were identified in each dataset. The predicted lines were selected based on predictions only and no field performance data was included.h

In 2021, the entire set of lines was reevaluated. The figure shows the performance of the entire population compared to the top and bottom 20% based on the different selection methods

The green box plots show the 2021 performance of the lines in the top 20% for each selection method. The red box plot shows the 2021 performance of the lines in the bottom 20% for each selection method.

Similar performance was observed with each selection method and all methods resulted in significant improvements compared to the entire population. This observation demonstrates that genomic predictions are as effective as a single-year yield trial in selecting the best lines to advance. This enables the breeding program to evaluate four times the number of lines, while maintaining the same size field program.