Fermentation (May 2021)

A Statistical Workflow to Evaluate the Modulation of Wine Metabolome and Its Contribution to the Sensory Attributes

  • José Manuel Muñoz-Redondo,
  • Belén Puertas,
  • Gema Pereira-Caro,
  • José Luis Ordóñez-Díaz,
  • María José Ruiz-Moreno,
  • Emma Cantos-Villar,
  • José Manuel Moreno-Rojas

DOI
https://doi.org/10.3390/fermentation7020072
Journal volume & issue
Vol. 7, no. 2
p. 72

Abstract

Read online

A data-processing and statistical analysis workflow was proposed to evaluate the metabolic changes and its contribution to the sensory characteristics of different wines. This workflow was applied to rosé wines from different fermentation strategies. The metabolome was acquired by means of two high-throughput techniques: gas chromatography–mass spectrometry (GC-MS) and liquid chromatography–mass spectrometry (LC-MS) for volatile and non-volatile metabolites, respectively, in an untargeted approach, while the sensory evaluation of the wines was performed by a trained panel. Wine volatile and non-volatile metabolites modulation was independently evaluated by means of partial least squares discriminant analysis (PLS-DA), obtaining potential markers of the fermentation strategies. Then, the complete metabolome was integrated by means of sparse generalised canonical correlation analysis discriminant analysis (sGCC-DA). This integrative approach revealed a high link between the volatile and non-volatile data, and additional potential metabolite markers of the fermentation strategies were found. Subsequently, the evaluation of the contribution of metabolome to the sensory characteristics of wines was carried out. First, the all-relevant metabolites affected by the different fermentation processes were selected using PLS-DA and random forest (RF). Each set of volatile and non-volatile metabolites selected was then related to the sensory attributes of the wines by means of partial least squares regression (PLSR). Finally, the relationships among the three datasets were complementary evaluated using regularised generalised canonical correlation analysis (RGCCA), revealing new correlations among metabolites and sensory data.

Keywords