Wine metrics

: revealing the volatile molecular feature responsible for the wine like aroma

Abstract

Wine represents a variety of aromas that stem from a complex, completely non-linear system of interactions among many hundreds of compounds. Aroma is usually associated with odorous or volatile compounds that result from fermentation. Yeast strain and fermentation conditions are claimed to be the most important factors influencing the aromas produced in wine. The fierce competition is forcing wine producers to understand better the expectations and preferences of their target market so they can produce wines accordingly. The motivation of this thesis is in the line with a tool for yeast strains selection according with the volatiles output giving a better acceptance by the consumers. This work provided insights about the connection between the wine fermentation metabolism and the product “wine-like” aroma perception. In this context were performed 3 replicas of 4 fermentations of a synthetic grape juice with a different S. cerevisae strain on each (3 wine strains: QA23, VL1, ZA and 1 cachaça strain: L328). The metabolic profiles of fermentations were obtained using gas chromatography attached to a flame ionization detector (GC-FID) or to mass spectrometry (GC-MS) and High Performance Liquid Chromatography (HPLC). A sensorial study also was used in order to evaluate the recognition of a “wine-like” aroma. This target approach coupled with unsupervised analysis, namely Singular Value Decomposition (SVD) and Hierarchical Cluster Analysis (HCA), revealed that the “wine-like” aroma key odorants are acetaldehyde, hexyl acetate and ethyl esters. L328 revealed to be the strain with better scores and correlation with the sensorial analysis scores of “wine-like”. A supervised analysis, Partial least squares regression (PLS-R) model, allowed the prediction (R2 = 0.8) of the “wine-like” scores of the samples during fermentation process. Finally, an untargeted metabolomic approach combining GC-MS data preprocessing with SVD was able to distinguish strains based on their metabolic profiles evolution during the fermentation time. L328 and QA23 strains revealed to be easily distinguished from each other and from the couple ZA and VL1. In conclusion this study demonstrated the potential of the use of chemometrics and bioinformatics approaches was explored in the characterization, prediction and classification of metabolic profiles from fermentations and the possibility of selection of the yeast strain according the final product characteristics.

Date of Award	8 Nov 2016
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	António Silva Ferreira (Supervisor)