Nontargeted screening based on EI-MS spectra using statistical methods: An investigative study of synthetic...

Mass spectrometry has evolved into a highly powerful tool for qualitative and quantitative chemical analyses. However, the identification of trace amounts of previously unknown structures in complex chemical matrix environments remains challenging. The rapid emergence of new synthetic cannabinoid substances is a typical example of this. Existing laboratory techniques are mostly based on methods used for lists of known illegal compounds. This situation poses a challenge to traditional data analysis and the risk of missing the compounds. Therefore, we propose to develop and validate a statistical model to classify newly emerging synthetic cannabinoid substances into a structural class or subclass.

We obtained 70 electrospray ionization spectra of indole/indazole synthetic cannabinoids from both the actual standard analysis and the SWGDRUG mass spectral library (version 3.10). Each sample consisted of 330 m/z variables and corresponding relative intensities. We first cleared the variables with a variance below 0.1. Principal component analysis (PCA) was performed on the variance-filtered data, and the two principal components were retained to generate new data for hierarchical clustering. After hierarchical clustering, we used the receiver operating characteristic method in this cluster.

Seventy synthetic indole/indazole cannabinoids were classified into four clusters. The side chain of cluster 1 is mainly fluorobenzyl, cluster 2 is pentyl, cluster 3 includes compounds from several structures, and cluster 4 is mainly fluoropentyl. The most relevant characteristic ions are m/z 109, m/z 252, and m/z 253 for cluster 1; m/z 144 and m/z 214 for cluster 2; and m/z 232 and m/z 233 for cluster 4.

This study provides a more objective and less time-consuming solution for characterizing synthetic cannabinoids. And this work validates the ability of PCA to extract characteristic fragment ions.