Electrochemical "Super-Fingerprinting" in Combination with Machine Learning for the On-Site Detection of Illicit Drugs.

Abstract

On-site multidrug sensing remains challenging due to the complexity of real samples and the differing detection requirements of individual substances. In the current study, we present successful electrochemical multidrug detection that overcomes these limitations by broadening the analytical framework, i.e., by performing square wave voltammetry simultaneously at four different conditions: pH 5, pH 7, pH 10/derivatizing, and pH 12. The combination of the four electrochemical fingerprints into a "super-fingerprint" was achieved by employing machine learning, specifically, the support vector machines algorithm coupled with principal component analysis. The proposed methodology was applied to the detection of cocaine, heroin, ketamine, amphetamine, methamphetamine, and MDMA as well as 24 adulterants/cutting agents. The novel detection technique demonstrated robust classification performance with very high specificity (∼90%), sensitivity (∼93%), and accuracy (∼92%), confirmed through the identification of the street samples of the six target drugs.