WineMetrics: A new approach to unveil the “wine-like aroma” chemical feature

Abstract

“The Human being has an excellent ability to detect and discriminate odors but typically has great difficulty in identifying specific odorants”(1). Furthermore, “from a cognitive point of view the mechanism used to judge wines is closer to pattern recognition than descriptive analysis.” Therefore, when one wants to reveal the volatile “wine-like feature” pattern recognition techniques are required. Sensomics is one of the most recent “omics”, i.e. a holistic perspective of a complex system, which deals with the description of substances originated from microorganism metabolism that are “active” to human senses (2). Depicting the relevant volatile fraction in wines has been an ongoing task in recent decades to which several research groups have allocated important resources. The most common strategy has been the “target approach” in order to identify the “key odorants” for a given wine varietal. That process produced an extensive list of substances that play, at least individually, a role on the perceived quality of the wine. However, the combined effect of volatiles responsible for triggering the mechanism of wine-like perception is less explored. A few works address that issue, using omission tests or tentative reconstruction of the wine aroma (3). While accepting that chemical reconstruction of the volatile ‘sensometabolome’ is an important branch of research in this area, our vision is that the reconstruction work should be transferred to “those who know better” i.e. the yeast. The absence of the impression substances feature description constitute an obstacle to define the role of the “aroma quality drivers” on a global market perspective, therefore we will attempt to reconstruct the chemical feature “driven” by the yeast. The objective of the present work was to perform comparative sensorial and metabolomics analysis with four yeast strains from different origins and/or technological applications (cachaça, wine and laboratory), during a fermentative process, in order to characterize their aroma profile and the ability to produce the “wine-like” aroma. Fermentations were analyzed daily by HS-SPME-GC-MS and submitted to sensory analysis. Multivariate tools such as principal component analysis (PCA) and partial least squares regression (PLS-R) were used in order to extract the compounds related with the “wine-like” aroma, by fusion of chemical with sensory data. This approach demonstrates that acetaldehyde; ethyl esters of fatty acids were related with “wine-like” aroma. With PLS-R we were able to develop a model capable to predict “wine-like” with a correlation of 0.8. With this methodology we were capable to create a pipeline that can be used in the future for strains selection which regards the ability to produce compounds related with the “wine-like” aroma.