Integration of Non-targeted Proteomics Mass Spectrometry with Machine Learning for Screening Cooked Beef Adulterated Samples
Keyuan Pu, Jiamin Qiu, Yongqi Tong, Bolin Liu, Zibin Cheng, Siyu Chen, Wen‐Xiu Ni, Yan Lin, Kwan‐Ming Ng
Abstract
The degradation of ingredients in heat-processed meat products makes their authentication challenging. In this study, protein profiles of raw beef, chicken, duck, pork, and binary simulated adulterated beef samples (chicken-beef, duck-beef, and pork-beef) and their heat-processed samples were obtained by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). Heat-stable characteristic proteins were found by screening the overlapping characteristic protein ion peaks of the raw and corresponding heat-processed samples, which were discovered by partial least-squares discriminant analysis. Based on the 36 heat-stable characteristic proteins, qualitative classification for the raw and heat-processed meats was achieved by extreme gradient boosting. Moreover, quantitative analysis via partial least squares regression was applied to determine the adulteration ratio of the simulated adulterated beef samples. The validity of the approach was confirmed by a blind test with the mean accuracy of 97.4%. The limit of detection and limit of quantification of this method were determined to be 5 and 8%, respectively, showing its practical aspect for the beef authentication.