Extra-virgin olive oil authentication: triacylglycerol profiling and machine learning using the Shimadzu MALDI-8020/MALDI-8030 and eMSTAT Solution

Significance of the Topic

Extra virgin olive oil (EVOO) is valued for its unique nutritional profile and high market price, making it a common target for adulteration with cheaper vegetable oils. Reliable and rapid authentication methods are essential to protect consumer health and maintain product integrity.

Objectives and Study Overview

This study aims to develop a fast, minimal-prep workflow combining matrix-free MALDI-TOF mass spectrometry and machine learning for the detection and classification of EVOO adulteration with sunflower oil. Simulated adulteration levels of 5, 10 and 20% were used to assess method sensitivity.

Methodology and Instrumentation

Sample Preparation and Adulteration

• Pure EVOO and sunflower oil samples were dissolved in chloroform.
• EVOO was spiked with sunflower oil at 5, 10 and 20% (v/v).
• Tricaprin served as an internal standard for mass alignment.

Instrumentation

• Shimadzu MALDI-8020/8030 benchtop linear MALDI-TOF in positive ion mode.
• Direct LDI (matrix-free) analysis on NaTFA-coated targets.
• Data acquisition and alignment via Shimadzu MALDI Solutions software.

Data Analysis

• Mass spectra processed with Shimadzu eMSTAT Solution.
• Principal component analysis (PCA) and RandomForest classification with KFold cross-validation.

Main Results and Discussion

Tag Profile Differences

• EVOO spectra featured dominant sodiated TAG peaks at m/z 881 and 907 (OPO/POO and OOO).
• Sunflower oil showed strong peaks at m/z 877, 901, 903 and 905 (LLL, LLO/LOL and OLO/OOL).
• Adulterated EVOO spectra revealed progressive shifts in TAG ratios and the emergence of the LLL marker (m/z 901) even at 5% adulteration.

Multivariate Analysis

• PCA score plots achieved clear separation of pure EVOO, adulterated EVOO and sunflower oil groups.
• PCA loading identified m/z 901 and 907 as key discriminators with high statistical weighting.
• Markers exhibited sensitivity and specificity above 97% and AUC values near unity.

Machine Learning Classification

• RandomForest model trained on 113 spectra and validated on 78 blinded samples.
• Achieved 100% accuracy in classifying pure and adulterated EVOO down to 5% sunflower content.

Benefits and Practical Applications

• Minimal solvent use and simple sample processing.
• Rapid, automated analysis with benchtop MALDI-TOF.
• High accuracy in detecting low-level adulteration, supporting quality control and regulatory compliance.

Future Trends and Potential Applications

• Extension to other edible oil matrices and broader lipid fingerprinting applications.
• Integration with advanced machine learning and cloud-based platforms for real-time authentication.
• Adoption in routine QC workflows to combat food fraud across the supply chain.

Conclusion

The combined use of matrix-free MALDI-TOF and machine learning provides a robust, high-throughput strategy for authenticating extra virgin olive oil and reliably detecting sunflower oil adulteration at low levels, offering a practical tool for industry and regulatory laboratories.

References

• No literature references were provided in the original document.