Guided by machine-learning models that predicted what would make Belgian beer taste more appealing, researchers tinkered with the golden nectar’s composition, greatly impressing blind taste testers. The method may lead to new recipes and improved tastes across a range of foods and beverages. Santé!
Controversial opinion incoming: I’ve never been a big beer drinker; I don’t like the taste. However, I understand – being Australian, I’m convinced it’s hard-wired into all of us from birth – that beer is considered a golden-hued nectar by many.
Understanding and predicting how we perceive and appreciate things we consume, such as beer, is a major challenge for the food and beverage industry. Surely, it makes good commercial sense for the industry to sway outliers like me to drink (or eat) its products. Well, researchers from KU Leuven, a university in Belgium, have developed a machine-learning model that could help develop beer flavors with greater consumer appeal, helping manufacturers to meet specific consumer needs.
“Predicting flavor and consumer appreciation from chemical composition is one of the ultimate goals of sensory science,” said the researchers. “A reliable, systematic and unbiased way to link chemical profiles to flavor and food appreciation would be a significant asset to the food and beverage industry.”
First, to generate a comprehensive dataset on beer flavor, the researchers selected 250 commercially available Belgian beers across 22 different beer styles. Most of the dataset comprised blonds (12.4%) and tripels (11.2%), reflecting their presence on the Belgian beer scene and the diversity of beers within these styles. Then, the researchers measured 226 different chemical properties for each beer, including brewing parameters such as alcohol content, pH, sugar concentration, and over 200 flavor compounds.
A trained tasting panel evaluated each of the 250 beers for 50 sensory attributes, including different hop, malt and yeast flavors, off-flavors and spices, which they scored. To expand upon the tasting panel’s data, the researchers collected 180,000 reviews of the 250 selected beers from the consumer review platform RateBeer. This provided numerical scores for appearance, aroma, taste, palate, overall quality and average overall score.
The researchers combined the chemical analyses, tasting panel assessments and public reviews and used them to train machine-learning models. They then leveraged the models to infer important contributors to sensory perception and consumer appreciation, cognizant that a product with low consumer appreciation doesn’t succeed commercially.
Ethyl acetate was identified as the most predictive parameter for beer appreciation. It typically conveys a fruity, solvent and alcoholic flavor. Ethanol, the most abundant beer compound after water, was the second most important parameter. In addition to directly contributing to beer flavor and mouthfeel, ethanol drastically influences the beverage’s physical properties, dictating how volatile compounds contribute to aroma. Lactic acid, which contributes to the tart taste of sour beers, was also rated highly. Interestingly, some of the most important predictive parameters were not well-established beer flavors and are commonly associated with negative beer quality. Ethyl phenyl acetate, for example, commonly linked to beer staling, was found to be a key factor contributing to beer appreciation.
Finally, the researchers tested whether their predictive models provided insight into beer appreciation. They specifically selected overall appreciation as the factor to be examined due to its complexity and commercial relevance. Because adding a single compound may produce a noticeable difference that unbalances a beer’s flavor profile, the researchers evaluated the effect of changing combinations of compounds. And because blond beers were strongly represented in the dataset, they selected a blond as the starting materials for these experiments.
Adjusting the concentrations of the most important predictors of overall appreciation – ethyl acetate, ethanol, lactic acid, and ethyl phenyl acetate – significantly improved overall appreciation, compared to controls, among a panel of trained tasters. The panelists noted an increased flavor intensity, sweetness, alcohol, and body fullness. To remove the contribution of ethanol to the results, a second experiment was run without adding ethanol. It resulted in a similar outcome, including increased overall appreciation. A further experiment tested whether the model’s predictions could improve appreciation for a non-alcoholic beer. Again, a mixture of predicted compounds – except ethanol – was added, resulting in a significant increase in appreciation, body, flavor and sweetness.
“Our study confirms that flavor compound concentration does not always correlate with perception, suggesting complex interactions that are often missed by more conventional statistics and simple models,” the researchers said. “The predictions of our final models, trained on review data, hold even for blind tastings with small groups of trained testers, as demonstrated by our ability to validate specific compounds as drivers of beer flavor and appreciation.”
The researchers are mindful of the societal burden caused by alcohol abuse and addiction and warn against using their method to exacerbate this.
“We encourage the use of our results for the production of healthier, tastier products, including novel and improved beverages with lower alcohol contents,” they said. “Furthermore, we strongly discourage the use of these technologies to improve the appreciation or addictive properties of harmful substances."
They hope that future studies will expand the scope of their research to include diverse markets and beer styles, identifying more drivers of appreciation.
“Soon, these tools may provide solutions in quality control and recipe development, as well as new approaches to sensory science and flavor research,” they said.
The study was published in the journal Nature Communications.
