Many variables affect the quality of the beer you enjoy. A beer's color depends on the malt used, while the aroma depends on the conditions of fermentation.
During fermentation, yeasts convert sugar into alcohol, and release carbon dioxide. This process transforms the wort—the unfermented mixture—and amino acids into aromatic compounds that determine the ultimate flavor of the beer.
Recently, one brewer set an important business goal: reducing fermentation time while preserving these aromas. The brewer turned to Damien Steyer, head of Twistaroma, an innovative startup firm that characterizes and infuses odorant volatile compounds and antioxidant in drinks; and Behnam Taidi, a biotechnology professor at the French engineering school CentraleSupélec. Together, they analyzed how concentrations of specific amino acids, alone or in combination, influence fermentation performance and beer flavor.
The researchers needed to identify the amino acids with the greatest influence on fermentation time and the production of the volatile compounds that influence the aroma. They used Minitab’s Design of Experiments (DOE) tools to plan experiments that could identify the most important factors with maximum efficiency. A designed experiment permits you to change more than a single variable for each experimental run. This minimizes the number of runs needed to obtain meaningful results and reach conclusions about how factors affect a response.
The team opted for a Plackett-Burman factorial design, which is capable of evaluating between 8 and 47 different factors and identifying which have the most impact on an outcome. This type of experiment is frequently called a “screening design,” because the main effects of each factor are confounded with two-factor interactions. When you have many possible factors, screening designs can tell you which are important, but it cannot provide details about how the important factors interact. But the screening experiment helps you avoid conducting unproductive experiments with factor that would not significantly affect the response.
To find out which amino acid concentrations were most important to the flavor of the brewer’s beer, the team carried out a Plackett-Burman design with 24 runs including a center point, which is an additional experimental run in which levels of all factors are set midway between the minimum and maximum values. Then they performed the experiments following the design, gathered the results data, and recorded it in Minitab for analysis.
When you analyze a designed experiment with Minitab, the output includes a visual graph called Pareto chart that helps you interpret results. By ordering the bars from largest to smallest, this bar chart can assist you in determining which factors comprise the "vital few" and which are the "trivial many." A cumulative percentage line helps you judge the added contribution of each factor. Any effect that extends beyond the line is statistically significant. The diagram below shows that lysine amino acid concentration has a significant effect on fermentation time.
When the results of the experimental runs are analyzed, lysine appears to be an influential amino acid. A regression model based on the data provides a R² of 90%, which means that 90% of the variation of the fermentation time is explained by the amino acids present in the model.
To refine their results, the team performed additional analysis to study the interactions between factors, and more thoroughly model the relationship between fermentation time and amino acid concentration. Steyer and Taidi used Minitab’s stepwise regression tool. Since they knew that Lysine was important, they told Minitab to keep that factor in each model. The software then automatically runs models by removing and adding each remaining amino acid one-by-one, which permits the researchers to check interactions and identify combinations of amino acids with a significant influence. The process systematically identifies a useful subset of predictors by adding the most significant variable or removing the least significant variable during each step.
Minitab offers three common procedures. Standard stepwise regression both adds and removes predictors as needed for each step. Forward selection starts with no predictors in the model, and Minitab adds the most significant variable for each step. Backwards elimination starts with all predictors in the model, and Minitab removes the least significant variable for each step.
If you select a standard stepwise regression, the terms you specify in the Model dialog box are candidates for the final model; eg.:
In the output of the regression analysis, Minitab provided an equation that can be used to calculate the time needed to reach fermentation depending on the concentration of the most influential amino acids. The team then applied the same approach to determine the effect of amino acids on the production of volatile compounds responsible for flavor.
Thanks to design of experiments and to stepwise regression techniques in Minitab, the team was able to determine which amino acids influence fermentation time and the production of volatile compounds. Their experiments showed that doubling the concentration of lysine can decrease fermentation time and influence the production of certain volatile compounds. Adding other specific amino acids also reduces the fermentation time, but causes a difference in beer aroma.
Since the concentrations of amino acids in the wort is linked to the cereals that make the wort, the results of this analysis will help the brewer select the best barley varieties according to their amino acid concentrations.
The brewer now has a scientific solution to beer design, and thanks to effective data analysis can meet both production constraints and consumers’ expectations.
Twistaroma and Chemical Engineering and Materials Lab at CentraleSupélec Engineering School
Increase fermentation performance maintaining beer flavor to meet brewer’s business objective.