Adulteration Detection of Multi-Species Vegetable Oils in Camellia Oil Using SICRIT-HRMS and Machine Learning Methods

Area of Science:

• Analytical Chemistry
• Food Science
• Spectrometry
• Machine Learning

Background:

• Camellia oil (CAO) adulteration with cheaper vegetable oils is a significant concern for food authenticity.
• Accurate and rapid detection methods are crucial for ensuring the quality and safety of edible oils.
• Existing methods may lack the sensitivity or throughput required for comprehensive adulteration analysis.

Purpose of the Study:

• To establish a rapid and precise method for identifying and quantifying multi-species vegetable oil adulterations in camellia oil (CAO).
• To evaluate the effectiveness of soft ionization by chemical reaction in transfer-high-resolution mass spectrometry (SICRIT-HRMS) coupled with machine learning algorithms.
• To provide a reference for developing non-targeted, high-throughput detection methods for edible oil authenticity.

Main Methods:

• Utilized SICRIT-HRMS to characterize the volatile profiles of pure and adulterated CAO samples.
• Employed various machine learning algorithms (CNN, RF, GBT, SVM, LR) for qualitative detection and quantitative prediction.
• Applied dimensionality reduction techniques (PCA, UMAP) alongside feature selection methods.

Main Results:

• SICRIT-HRMS effectively characterized volatile profiles, with the low m/z region (100-300) being important for classification.
• Machine learning models achieved high accuracies (98.70-100.00%) for qualitative detection and robust performance in multiclass classification (RF: 96.25-99.45%).
• The PCA-CNN model demonstrated optimal quantitative prediction of adulteration levels, showing high coefficients of determination and low error rates for olive oil (OLO) and sunflower oil (SUO).

Conclusions:

• The combination of SICRIT-HRMS and machine learning provides a rapid and accurate solution for identifying and quantifying multi-species vegetable oil adulterations in CAO.
• This approach offers a reliable tool for ensuring edible oil authenticity and combating fraudulent practices.
• The developed method serves as a valuable reference for future high-throughput, non-targeted edible oil analysis.