Mineral profiles and machine learning target identification of the ten Beaujolais crus

A mineral fingerprint for wine origin

A new study investigates whether mineral-element analysis and machine learning can distinguish wines from the ten Beaujolais crus. Located in France’s Auvergne-Rhône-Alpes region, the Beaujolais vineyard is internationally known for its geological diversity and the strong identities associated with its ten crus.

At the center of the research is a method introduced by the authors as the Mineral Wine Profile, or MWP. It is a high-throughput ICP-MS protocol designed to provide semi-quantitative measurements of more than 40 mineral elements. The approach is intended to turn complex elemental profiles found in wine into patterns that can support identification of geographical origin.

Gamay wines from two vintages

The study includes 1,288 Gamay wines from the 2022 and 2023 vintages. The samples were collected across the ten Beaujolais crus and analyzed using the MWP protocol. The research then integrates these mineral profiles with machine learning to discriminate among the individual crus.

The supplied excerpt and summaries do not specify which machine-learning algorithms were used, how the dataset was divided for training and testing, or what classification accuracy was achieved. It is therefore not possible from the available material to assess how reliably the system identified each origin or whether its performance remained consistent across the two vintages.

Authenticity and appellation protection

The practical motivation for the work is the protection of appellation authenticity. Mislabeling and fraud can reduce economic value for producers while also weakening consumer trust. A scalable combination of chemical analysis and machine learning could therefore offer an additional tool for checking claims about where a wine was produced.

However, the material provided does not report the cost of the protocol, its reproducibility between laboratories, or its suitability for use in regulatory procedures. Those factors would require further assessment before the approach could be evaluated for broader operational adoption. The available information also does not establish whether the method can be transferred to other grape varieties, wine regions or vintages beyond the dataset examined here.

The original study was published by arXiv – Viticulture & Grapevine.