Deep Learning-Based Prediction and Identification of Synthetic Cannabinoid Spectrum

Synthetic cannabinoids are a new class of synthetic psychoactive substances. By slightly modifying their chemical structures, new types of compounds could be derived, and it was precisely by taking advantage of this feature that unscrupulous elements were able to evade legal regulation through structural modification. However, mass spectrometry, as a mainstream detection instrument, was found to have obvious limitations in the coverage of its mass spectrometry libraries, which were constructed by relying on data from actual tests and information on known compounds in commercial databases. Due to the difficulty in obtaining standards for novel synthetic cannabinoids, the data in the mass spectrometry library could not be rapidly updated. To address the above problems, a deep learning-based mass spectrometry library prediction and expansion method was proposed in this study, combined with a similarity search algorithm, aiming to achieve the rapid prediction of mass spectrometry data for novel synthetic cannabinoids and the quick identification of novel synthetic cannabinoids.