Artificial intelligence application designed to screen for new psychoactive drugs based on their ATR-FTIR spectra

The analysis of the pharmacological effects of new psychoactive drugs reported by the United Nations Office on Drugs and Crime (UNODC) up to December 2016 indicates that most of these illicit drugs were stimulants, synthetic cannabinoids and hallucinogenic amphetamines. The emergence of these substances has been associated with severe intoxications, including an important number of fatalities. Hence, fast and reliable analytical tool able to detect new compounds belonging to these classes of drugs of abuse are highly needed. We are presenting a comparative study of several artificial intelligence applications designed to recognize the class identity of hallucinogenic amphetamines and of synthetic cannabinoids based on their spectra obtained by Attenuated Total Reflection – Fourier Transform Infrared Spectroscopy (ATR-FTIR). A first application has been built by using Partial Least Squares Regression applied to the most relevant absorptions selected by using a genetic algorithm. Then, the K-Nearest Neighbors and Random Forest procedures have also been tested. The results have been compared from the point of view of the classification rates. Taking into account the legal implication of the analytical results of these forensic applications, a special attention has been paid to the assessment of their sensitivity and selectivity. The limits of these class identity assignment systems are also discussed.