Penn State Scientists Uncover Oak Tannins That Shape Wine Flavor Using Machine Learning

Researchers at Penn State University have developed a new method to identify and chemically characterize the specific tannins from oak barrels that contribute to the flavor and mouthfeel of wine. For centuries, winemakers have aged wine in oak barrels, which are known to impart flavors such as vanilla, spice, caramel, and smoke. However, until now, scientists have struggled to pinpoint which tannins—compounds found in both grape skins and oak wood—are responsible for these distinct tastes.

The research team, led by Misha Kwasniewski, associate research professor of food science at Penn State’s College of Agricultural Sciences, used advanced chemical analysis combined with machine learning to “fingerprint” individual tannins in wine. Their findings were published in the journal Food Chemistry. The new approach allows scientists to separate complex mixtures of tannins, identify their molecular components, and quantify them even at low concentrations.

Traditional methods for identifying tannins involved breaking them apart with acid, which often damaged the compounds before they could be accurately measured. The Penn State team instead used a technique called in-source fragmentation with a mass spectrometer. This tool breaks tannins into molecular fragments and detects the unique chemical signature of each fragment. These signatures form a fingerprint that can distinguish between tannins from grapes and those from oak.

Because tannins can combine in billions of ways, the researchers turned to machine learning—a form of artificial intelligence—to sort through the complex data and identify which combinations are present in different wines. Kwasniewski’s lab has developed a machine learning model that continues to expand its ability to recognize new classes of tannins. The goal is to better understand how these compounds influence flavor, bioactivity, and plant biochemistry.

The study focused on hydrolysable tannins—those that can be broken down by water—which are found in oak wood and transferred into wine during barrel aging. These include ellagitannins, common in oak and influential in wine aging and structure, and gallotannins, found in both wood and grape skins, which contribute to bitterness and astringency.

The research was conducted in collaboration with Gallo, a California-based winery that provided some of the 22 red and 20 white wines analyzed. The team also experimented with 15 commercial oak chip products representing French, Hungarian, and American oak—the three main types used for wine barrels. By adding these chips to wine samples, they could compare how different oaks affected the levels of hydrolysable tannins.

Results showed that French oak contained the highest concentrations of both ellagitannins and gallotannins, followed by Hungarian oak and then American oak. The differences among oak types are likely due to their botanical origins, native tannin content, and cooperage practices such as seasoning the wood or bending staves with fire or steam.

The researchers also studied how toasting—the process of heating the inside of barrels—changes the chemical makeup of the wood. Toasting breaks down harsh tannins into softer compounds and caramelizes sugars in the wood, which alters the sensory properties imparted to wine.

Yanxin Lin, who recently earned a doctorate in food science from Penn State and is now a postdoctoral scholar at UC Davis, was first author on the study. Other contributors included Bruce Pan, Robert “Qiang” Sui, and Ping Yu from Gallo. The project received support from the Crouch Endowment for Viticulture, Enology, and Pomology Research at Penn State’s College of Agricultural Science, the American Vineyard Foundation, and the U.S. Department of Agriculture’s National Institute of Food and Agriculture.

This research marks a significant step toward understanding how specific compounds from both grapes and oak barrels shape the taste profile of wine. By using machine learning to analyze complex chemical fingerprints, scientists hope to give winemakers more control over flavor development during aging.