Development of an integrated searching system for rapid and accurate classification of edible oils by matrix-assisted laser desorption/ionization mass spectrometry

Background: Edible oil fraud has been frequently reported; however, the widely adopted analytical method for edible oils requires a lengthy column separation and skilled technicians to manually analyze the sophisticated fatty acid profile of edible oil samples for their authenticity verification. Although the MALDI-MS could improve the efficiency of edible oil analysis, it relies on offline manual data analysis, limiting its high-throughput capability. Therefore, a simple automated system integrated with the MALDI-MS is necessary for high-throughput online edible oil analysis to safeguard food safety from edible oil fraud. Results: The Gaussian mixture model – logistic regression – decision tree (GMM-LR-DT) model was established in this study for rapid and accurate classification of edible oils with an expanded spectral database based on MALDI-MS. Overall, the model achieved the average accuracy, sensitivity, precision, and F1-score of 0.997, 0.986, 0.998, and 0.991 in the 5-fold cross-validation, and 0.997, 0.998, 0.985, and 0.991 in the validation, respectively, with the scoring system reflecting the confidence levels in the classification of edible oil samples. The GMM-LR-DT model was further advanced into a one-step and straightforward program, called OilAnalysis, which was directly integrated with the equipment software for the analysis of MALDI-MS spectra, achieving high-throughput online analysis of edible oils using MALDI-MS. 95 % (133 out of 140) of the testing samples obtained from a separated MALDI-MS were correctly classified using the OilAnalysis with a spectrum processing speed of 1.7 spectra per second. Significance: A novel streamlined online system, OilAnalysis, integrated with the software for MALDI mass spectral analysis was developed. OilAnalysis coupled with the direct sample loading protocol and automatic spectral acquisition is able to achieve rapid and accurate classification of large amounts of edible oil samples, preventing edible oil fraud and safeguarding food safety in real life.