Machine learning supported olive oil compound profiling for assessing geographic and cultivar authenticity

Area of Science:

• Food Chemistry
• Analytical Chemistry
• Chemometrics

Background:

• Virgin Olive Oil (VOO) authenticity and geographical origin are crucial quality parameters.
• Milling processes can introduce variability, confounding the identification of genuine markers.
• Accurate authentication requires methods that can distinguish cultivar and geographical influences from processing effects.

Purpose of the Study:

• To develop an integrated metabolomics and machine learning (ML) framework for robust VOO authenticity assessment.
• To identify reliable chemical markers of Tuscan VOO origin, independent of milling variations.
• To predict sensory attributes of VOO using chemical profiles.

Main Methods:

• A multi-platform analytical workflow combining 1H NMR, HS-SPME GC-MS, and HPLC-DAD-FLD.
• Quantitative data from analytical platforms were integrated to train ML models (Random Forest, TreeNet, MARS).
• A two-stage ML approach was used to remove mill-related variance and identify geographical markers.

Main Results:

• Accurate geographical classification of Tuscan VOOs with ~80% cross-validated accuracy.
• Identification of key chemical markers including secoiridoids, sterols, terpenoids, and lipid derivatives.
• High predictive performance (R² > 0.83) for sensory attributes like artichoke, fruity, and rancid notes.
• Discovery of margaric acid as a marker associated with specific olive cultivars.

Conclusions:

• The integrated metabolomics and ML framework effectively identifies geographical markers and predicts sensory attributes.
• This approach minimizes confounding factors from milling, enhancing the reliability of VOO authentication.
• The study provides a foundation for advanced VOO quality assessment and regional characterization based on chemical profiles.