Optimisation of quality features of new wheat beers fermented through sequential inoculation of non-Saccharomyces and Saccharomyces yeasts

The choice of the starchy ingredients as well as that of the yeasts strongly can represent a useful tool to differentiate the final beers. Our research investigated twelve white beers obtained applying a 2-factor mixed 3-level/4-level experimental design. The first factor was the cereal mixture, with 3 combinations of barley malt (65 %) and unmalted wheat (35 % of common, durum, or emmer). The second factor was the yeast used to carry out the fermentation trials, i.e.: a S. cerevisiae starter strain (WB06); an oenological S. cerevisiae strain (9502); two mixed starters made of an oenological Schizosaccharomyces pombe strain (6956) and, alternatively, one of the two S. cerevisiae strains. Most beer attributes were significantly (p < 0.05) influenced by the two considered factors with the following exceptions: the wheat species did not affect maltotriose, maltose, pH, total and volatile acidity, floral flavour, and sweetness; the yeast did not exert significant effects on foam colour, turbidity, overall olfactory intensity, yeast flavour, and body. The flavour of fruits and aromatic herbs were not influenced by the factors studied. Alcohol content was maximised using the unmalted durum wheat (~7 %) and S. cerevisiae WB06 (~6.8 %). The beer antioxidant content was increased by the use of emmer (566 mg/L) and by the application of the mixed inoculum (478–487 mg/L). The beers made with unmalted common wheat and fermented by the S. cerevisiae strains alone obtained the best overall sensory score (3.7). As shown by the Principal Component Analysis, the beers were better classified by the type of unmalted wheat than by the fermenting yeast. A multiple regression analysis was performed by fitting the analytical parameters that highlighted significant differences among the beers to a second-order polynomial model. Data concerning colour, glycerol concentration, FC-TPC, and antioxidant activity were satisfactorily predicted (R2 > 0.8) by the fitted models.