Portable mass spectrometry and machine learning for detecting Angelica sinensis adulteration

Angelica sinensis Radix (ASR), commonly known as Dong quai, is a nutrient-rich edible herb widely used in functional foods and dietary supplements. Due to its high market value and the morphological similarity of related species, unscrupulous vendors frequently adulterate ASR with Angelica pubescens, Notopterygium incisum, and Ligusticum chuanxiong. This not only affects product quality but also poses potential food safety risks. Thus, a rapid and reliable method for detecting ASR adulteration is crucial for food quality control and consumer safety. In this study, a portable mass spectrometer (PMS) combined with chemometric modeling was employed for the rapid detection of adulteration in ASR. After appropriate preprocessing and feature selection, machine learning models including random forest, k-nearest neighbors, and a snake optimization-support vector machine (SO-SVM) were developed. Among them, the SO-SVM model based on features selected by the least absolute shrinkage and selection operator achieved the highest classification accuracy of 0.99. In addition, through the variable importance analysis calculated by the random forest model, the ions with m/z of 132.56, 114.42, and 164.58 were identified as the most critical feature variables for authenticity identification. Overall, these results demonstrate that PMS, combined with machine learning, enables accurate and rapid discrimination between authentic and adulterated ASR samples. It offers a simple, fast, and reliable solution for on-site authenticity detection and quality assurance in the food industry.