A machine learning-based nuclear magnetic resonance profiling model to authenticate ’Jerez-Xérès-Sherry’ wines

Abstract

Traditionally, wine quality and certification have been assessed through sensory analysis by trained tasters. However, this method has the limitation of relying on highly specialized individuals who are typically trained to evaluate only specific types of products, such as those associated with a particular Denomination of Origin (D.O.), etc. While tasters can often identify instances of fraud, they are generally unable to pinpoint its origins or explain the mechanisms behind it.

On the other hand, classical biochemistry has made significant progress in understanding various aspects of winemaking. However, it has yet to identify the specific metabolites responsible for the unique characteristics of wines, particularly those influenced by complex variables involving multiple compounds, such as geographical differences between regions or vineyards. The concept of the “Terroir fingerprint” has emerged as a novel approach to wine certification. The concept refers to the unique characteristics imparted to a wine by its geography, climate, and aging process. Nuclear Magnetic Resonance (NMR) technology plays a pivotal role in establishing this “Terroir fingerprint” because it enables precise identification, quantification, and differentiation of the compounds present in wine. NMR provides a highly reproducible and specific method for certification. This work introduces an innovative project that combines NMR technology with Artificial Intelligence to create a profiling model for certifying the authenticity and quality of ‘Jerez-Xérès-Sherry’ wines.