Traceability and discrimination of opium poppy shell analogues using HS-GC-IMS combined with machine learning algorithms.

Abstract

Illegal adulteration has been a critical issue in food safety, emerging as a focal point in forensic science. This situation has led to an increased demand for effective detection and monitoring technologies. Opium poppy shells are a critical source of drugs, and the accurate tracing and identification of their analogues are essential in drug-related cases. The features of volatile compounds in six opium poppy shell analogues (OPSA) were characterized using headspace-gas chromatography-ion mobility spectrometry (HS-GC-IMS) in this study, and an accurate model for origin tracing was established through the integration of machine learning algorithms. A total of 213 volatile compounds were accurately identified, with esters, ketones, aldehydes, alcohols, and alkenes being the most abundant compounds. Additionally, two supervised machine learning algorithm classification models were established based on the HS-GC-IMS dataset to predict the categories of OPSA, including the orthogonal partial least squares discriminant analysis (OPLS-DA) and random forest models, and were subsequently compared with unsupervised models. By employing the random forest classification model, significant volatile compound characteristics were recognized, resulting in enhanced efficiency. Furthermore, the model achieved an out-of-bag (OOB) error value of 0, indicating excellent predictive capability for tracing and distinguishing OPSA. Our research findings indicate that the integration of HS-GC-IMS with machine learning is expected to enhance the efficiency and accuracy of tracing and identifying the categories of OPSA, thereby providing theoretical support for litigation and judicial processes.