Support Vector Machine Classification of Adulterated Illicit Opioids Using Paper-Spray Mass Spectrometry

The complexity and potency of the illicit opioid supply in North America has become increasingly concerning for people who use drugs. Drug checking efforts aim to keep up with the evolving psychoactive components present in the illicit drug supply. While targeted paper-spray mass spectrometry (PS-MS) methods are effective for trace detection and quantitation, they are limited in their ability to detect emerging substances. Here, we demonstrate the use of a support vector machine (SVM) classifier to detect opioid samples containing an additional component outside of a routine targeted analysis method, using ortho-methylfentanyl as proof of concept. This approach allows for the focused selection of samples for additional screening to identify emerging adulterants from full-scan data collected during a 2 min PS-MS analysis. The developed classifier achieved a precision of 0.77 and a recall of 0.94 for the detection of ortho-methylfentanyl in samples containing fentanyl and caffeine. Shapely additive explanations were used to explain and interpret the results of the developed SVM classifier, and enabled a better understanding of the composition of illicit opioids. Our work demonstrates a new and effective approach for identifying adulterants from unit mass resolution mass spectrometry data generated during routine on-site drug analysis.