Screening unknown novel psychoactive substances using GC–MS based machine learning

• Published in: Forensic chemistry Vol. 34; p. 100499
• Main Authors: Wong, Swee Liang, Ng, Li Teng, Tan, Justin, Pan, Jonathan
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2023
• Subjects: Artificial intelligence; Chemometrics; Gas chromatography mass spectrometry; Machine learning; Neural network; Random forest; Gas chromatography mass spectrometry; Random forest; Neural network; Artificial intelligence; Chemometrics; Machine learning
• ISSN: 2468-1709; 2468-1709
• DOI: 10.1016/j.forc.2023.100499

We developed machine learning models which uses mass spectra to predict class of unknown NPS, with macro-F1 scores of 0.9 and improved accuracies over database matching. [Display omitted] •New structures of Novel Psychoactive Substances enter illegal market constantly.•Databases may not be updated in time, enabling these drugs to evade detection.•Machine learning can detect these substances using an incomplete database.•Chemical isomers have to be considered when doing performance benchmarking. In recent years, there is a large increase in structural diversity of novel psychoactive substances (NPS), exacerbating drug abuse issues as these variants evade classical detection methods such as spectral library matching. Gas chromatography mass spectrometry (GC–MS) is commonly used to identify these NPS. To tackle this issue, machine learning models are developed to address the analytical challenge of identifying unknown NPS, using only GC–MS data. 891 GC–MS spectra are used to train and evaluate multiple supervised machine learning classifiers, namely artificial neural network (ANN), convolutional neural network (CNN) and balanced random forest (BRF). 7 classes, comprising 6 NPS classes (cathinone, cannabinoids, phenethylamine, piperazine, tryptamines and fentanyl) and other unrelated compounds can be effectively classified with a macro-F1 score of 0.9, averaged across all cross-validation folds. These results indicate that machine learning models are a promising complement as an effective NPS detection tool.