Identification and classification of cathinone unknowns by statistical analysis processing of direct analysis in real...

An approach to the rapid determination of the structures of novel synthetic cathinone designer drugs, also known as bath salts, is reported. While cathinones fragment so extensively by electron impact mass spectrometry that their mass spectra often cannot be used to identify the structure, collision-induced dissociation (CID) direct analysis in real time-high resolution mass spectrometry (DART-HRMS) experiments furnished spectra that provided diagnostic fragmentation patterns for the analyzed cathinones. From this data, neutral loss spectra, which reflect the presence of specific chemical moieties, could be acquired. These spectra showed striking similarities between cathinones sharing structural features such as pyrrolidine rings and methylenedioxy moieties. Principle component analysis (PCA) of the neutral loss spectra of nine synthetic cathinones of various types including ethcathinones, those containing a methylenedioxy moiety appended to the benzene ring, and pyrrolidine-containing structures, illustrated that cathinones falling within the same class clustered together and could be distinguished from those of other classes. Furthermore, hierarchical clustering analysis of the neutral loss data of a model set derived from 44 synthetic cathinones, furnished a dendrogram in which structurally similar cathinones clustered together. The ability of this model system to facilitate structure determination was tested using 4-fluoroethcathinone, 3,4-methylenedioxy-α-pyrrolidinohexanophenone (MDPHP), and ethylone, which fall into the ethcathinone, pyrrolidine-containing, and methylenedioxy-containing subclasses respectively. The results showed that their neutral loss spectra correctly fell within the ethcathinone, pyrrolidine-containing and methylenedioxy-containing cathinone clades of the dendrogram, and that the neutral loss information could be used to infer the structures of these compounds. The analysis and data processing steps are rapid and samples can be analyzed in their native form without any sample processing steps. The robustness of the dendrogram dataset can be readily increased by continued addition of newly discovered structures. The approach can be broadly applied to structure determination of unknowns, and would be particularly useful for analyses where sample amounts are limited.