High accuracy machine learning identification of fentanyl-relevant molecular compound classification via constituent functional group analysis

Abstract

Fentanyl is an anesthetic with a high bioavailability and is the leading cause of drug overdose death in the U.S. Fentanyl and its derivatives have a low lethal dose and street drugs which contain such compounds may lead to death of the user and simultaneously pose hazards for first responders. Rapid identification methods of both known and emerging opioid fentanyl substances is crucial. In this effort, machine learning (ML) is applied in a systematic manner to identify fentanyl-related functional groups in such compounds based on their observed spectral properties. In our study, accurate infrared (IR) spectra of common organic molecules which contain functional groups that are constituents of fentanyl is determined by investigating the structure–property relationship. The average accuracy rate of correctly identifying the functional groups of interest is 92.5% on our testing data. All the IR spectra of 632 organic molecules are from National Institute of Standards and Technology (NIST) database as the training set and are assessed. Results from this work will provide Artificial Intelligence (AI) based tools and algorithms increased confidence, which serves as a basis to detect fentanyl and its derivatives.