As a peanut breeder, Jeff Dunne is concerned that the flavor of roasted peanuts has declined over the past 20 years. He said steps must be taken to improve the flavor profile of the Virginia-type peanuts his program breeds at North Carolina State University.

NC State’s study analyzed 17 years of breeding data and found a trend: Roast peanut flavor — the main attribute associated with a peanut’s taste — has declined over time. Dunne reported these findings at November’s Emerging Research Showcase at the North Carolina Biotechnology Center in Research Triangle Park.

Dunne said plant breeding methods along with machine learning can improve the flavor profile of the Virginia-type cultivars without harming disease resistance and high-yield characteristics.

In 2025, NC State will launch a research initiative — supported by industry partners Mars Wrigley, Qualy Sense and JLA — to use machine learning in its applied breeding program to improve peanut flavor.

Tools will include a QSorter to evaluate raw peanut kernels, and a hand-held sensor to evaluate the roasted peanut paste sample to provide a spectra of the peanut lines in the breeding program. For this project, JLA will provide the flavor testing panel as part of its working agreement with Mars Wrigley.

Related:Analysis: Peanut farmers annually losing revenue

“Basically, what we are going to do is build some machine learning models on an initial base population, and then we are going to apply those models to our Bailey II variety with Chimera and NC Bunch populations,” Dunne said. “What that will effectively do is allow us to now select within the confines of our actual testing program, as opposed to waiting to the very end.”

The Virginia-type accessions, Chimera and NC Bunch, were selected because of their flavor attributes. Both are known for their good roast peanut flavor. Bailey II was selected because it is the most widely planted Virginia-type cultivar.

Finding better flavor

NC State has been evaluating roasted peanut flavor since 1975 when the first descriptive sensory panel for peanut flavor was established as part of the university’s peanut breeding program. The panel will add terms for off flavors and changes in words used for flavor descriptions.

These terms include roast peanut, sweet aromatic, dark roast, raw bean, woody-hull skins, cardboard, earthy, painty, plastic chemical, metallic, fruity fermented, spice (cinnamon, nutmeg), sweet, sour, salty, bitter, astringent, tongue or throat burn, and ashy.

Dunne said machine learning will streamline data collection to more efficiently select for flavor improvement, focusing on the roast peanut flavor attribute.

Related:Peanut farmers answer for what and why they do

While roast peanut flavor is declining, Dunne noted that other flavor categories such as sweetness and sweet aromatic are improving. The challenge is to improve roast peanut flavor that companies such as Mars Wrigley are demanding.

“We’re reducing cardboardy, off-type flavors. But we really want the robustness of that roast peanut flavor score,” Dunne said. “What do we do from a breeding program standpoint? We start scouring for germplasm that actually has improved roast peanut flavor scores.”