(218i) Identifying Novel Fentanyl Analogs from Mass Spectral Measurements

Illegal designer drugs are a rampant societal problem. Clandestine laboratories produce new drugs with molecular structures that are sufficiently different from currently scheduled drugs, which results in the drugs being more easily distributed before law enforcement can act. Fentanyl related compounds (FRCs) are one major class of designer drugs. We recently developed the NIST Fentanyl Classifier to accelerate the identification process of new FRCs. This method combines an evaluated mass spectral library of known fentanyl analogs, a new chemistry-informed feature similarity measure called hybrid similarity, and traditional clustering techniques to determine whether a compound is an FRC given its electron ionization mass spectrum, and produces a probable molecular structure if the compound is deemed sufficiently related to fentanyl. Using the NIST Fentanyl Classifier, we correctly propose structures for most FRCs that differed from fentanyl by the addition of one or two moieties. We incorrectly classified a limited set of “unrelated” compounds as FRCs, but these compounds only differed from fentanyl by the loss of a single chemical moiety. In this presentation, we cover the measurements and algorithms that underlie the NIST Fentanyl Classifier, focusing on how leveraging the predictable fragmentation of fentanyl for feature selection greatly simplified the problem. We will also discuss similar approaches and extensions to other chemical applications.