Simplifying, Greening & Improving the Throughput of Edible Oil Testing & Reporting

Justification: As the food industry evolves, so too does the landscape for testing. Producers and consumers are increasingly concerned with safety, quality, authenticity, origin, and verification of claims (e.g., halal, kosher, vegetarian, vegan, low- fat, gluten-free, non-GMO, organic, extra virgin, …). Although testing methods are available to address these concerns, many are too complex, costly, cumbersome and slow.

Objective: To combine two direct mass spectrometric (MS) platforms (i.e., no inline liquid chromatography) with machine learning (ML) tools to develop and evaluate a practical approach to extra virgin olive oil (EVOO) testing.

Methods: Triglycerides (TAGs) fingerprints and small molecule fingerprints (i.e., polyphenols and other components) were generated on over 100 EVOO samples using matrix-assisted laser desorption/ionization (MALDI) MS and direct analysis in real time (DART) MS. Specific phenols including tyrosol, hydroxytyrosol, oleuropein aglycone, oleocanthal and oleacein were quantified by DART-MS/MS in the multiple reaction monitoring mode. ML models were developed and evaluated for assessing three specific oil attributes: oil type/authenticity, oil grade/quality and freshness.


Results: Performance evaluation of our ML models (i.e., 10 repeats; 5-fold cross validation) are shown in ROC curves using a support vector machine algorithm. Area under the curve, mean values of balanced accuracy, F1 scores, and SDs are also included.

Conclusions: Integrating ML with direct MS allows targeted characterization of specific product traits. We are exploring the inclusion of additional MS approaches (e.g., for measuring volatile and/or elements and their isotopes) and the development of additional models to assess other edible oil characteristics including sensory defects, botanical origin(s), country of origin, nutritional content, and contaminant/additive detection. Our workflow is fast, reduces human error, and can be fully automated. As consumers become more informed and engaged, there’s growing interest in providing clear, easily interpretable information that ensures food safety and builds trust with consumers.