Written by Carla Espinoza Gutierrez
The quest for the perfect peanut has found a new companion. Researchers at the University of Georgia (UGA) are bringing major technological advances to breeders with a pioneering method. A machine learning model that acts as a super-efficient data-driven “taste tester.”
Flavor, a trait that largely determines success in the snack aisle, has historically been a bit of a guessing game in agricultural breeding, as even tried-and-true varieties can be inconsistent. That’s where Joon Hyuk Seo’s team comes in.
“Flavor evaluation is currently one of the most time-consuming and expensive bottlenecks in peanut breeding,” said Suh, assistant professor in the UGA School of Food Science and Technology. While human sensory panels remain the gold standard, Hsu explained: FreshFruitPortal.com they are Requires large amounts of samples that are typically not available until late in the breeding pipeline— often after years of choosing a field.
Photo by Caroline Newbern | Uga
a Multidisciplinary collaborationSuh’s project promises: Reduce development costs and accelerate delivery of consistent flavor peanut varietiesfundamentally changing the way one of the state’s most profitable crops is grown.
“Researchers are often imagined as working alone in a lab, but this project is the opposite,” Lee said of the model’s development. “Our team is united, Food chemistry, sensory science, machine learning, peanut breedingan area that may not seem like an obvious partner. ”
Thanks to this integrated approach, This model can predict the complete flavor profile of roasted peanuts Ahead of traditional methods, Analyze the exact chemical fingerprints in raw peanuts.
The goal is Provides breeders with a faster way to screen hundreds of experimental peanut lines Early stages of development, when only small sample sizes are available.
According to the U.S. Department of Agriculture, approximately Peanut production in 2025 will be 6.1 billion poundsGeorgia accounts for about half of the nation’s production.
Roasting peanut tasting based on data
Environmental conditions, genetics, maturity, and postharvest handling all affect the chemical profile of peanuts before roasting begins, further complicating efforts to predict flavor outcomes.
“The biggest challenge is that peanut flavor is not the product of one or two compounds, but a complex mixture of compounds and their relative levels,” Suh explained.
Suh said the team is using Identify patterns using chemical profiling with sensory panel data and machine learning Leads to a desirable roasted peanut flavor.
Photo credit: UGA
The project has already identified pyrazines, a class of natural or synthetic nitrogen-containing compounds associated with roasted, nutty flavors, as well as sugars and amino acids that contribute to the development of flavor precursors. Sue said. The team continues to study which compounds serve as the most reliable predictive markers. Beyond varieties and growing seasons.
Namhee Lee, a graduate researcher in Suh’s lab who led much of the analysis and modeling work on the project, said one of the most challenging parts of the project was: make perfect of Extraction methods to identify and track flavor markers.
“The aroma of roasted peanuts comes from compounds that disappear quickly once released, so temperature, timing, and conditions must be carefully adjusted to catch the aroma before they disappear,” she emphasized.
This study also suggests the possibility of predicting the flavor of roasted peanuts from raw peanut data before roasting.
“If that happens, screening could become even faster and easier. No need to roast all samples before analysis” said Lee.
These discoveries can directly impact decisions made in the field, and Lee said this contribution is very meaningful to her on a personal level.
“The most meaningful thing for me personally was realizing that my work as a food chemist could directly contribute to peanut breeding. Seeing that connection made me feel really proud as a scientist.” she said.
Decoding the flavor code
Another important feature of this project is that Perfect accuracy is not required for early-stage screening for a model to be commercially usefulexplained Mr. Xu.
“Most important is the ability to reliably ‘rank and discriminate’ between breeding lines,” he says.
Through research, it is possible to Helping breeders eliminate low-potential varieties before committing years of resources to field development and sensory testing.
“This could give breeders and researchers a faster, more objective and science-based way to compare flavor differences between varieties at an early stage when sensory evaluation is not yet practical,” Lee said.
Suh described the framework as follows: The program could eventually expand beyond peanuts to other crops and specialty crop breeding programs. Flavor may still be difficult to measure in the early stages of development.
“We believe our framework is widely transferable,” Hsu said. “Crops and fruits where flavor is a complex function of many compounds and where sample volumes early in breeding are too small for traditional sensory evaluation are good candidates.”
*All photos without credits are for reference only.
Related articles
Peanut Board debuts first-ever unified industry brand
