Electrochemical fingerprinting combined with machine learning algorithm for closely related medicinal plant identification

Abstract

Medicinal plants have been widely used in the treatment of various diseases for human health. We developed a novel method for the identification of closely related medicinal plants using a machine learning (ML)-based electrochemical fingerprinting platform. Firstly, the system featured a bare glassy carbon electrode capable of recording the voltammetric response of active components in medicinal plants as electrochemical fingerprints. Subsequently, different algorithms and various datasets were employed to analyze the correlation between the above electrochemical fingerprint data and the medicinal plant species. As a proof-of-concept, 6 species of Anoectochilus roxburghii (A. roxburghii) were selected as the verification samples. The electrochemical fingerprints of the samples were measured by differential pulse voltammetry in two buffer solutions. Thereafter, four powerful ML algorithms were utilized for the identification of A. roxburghii with different datasets. The results showed that the accuracy of identifying species reached 94.4 % by the nonlinear support vector machines based on the slope data of electrochemical responses in two buffer solutions, evidencing the successful discrimination of closely related medical plants by this method. Additionally, ML combined with electrochemical fingerprinting approaches had the advantages of being rapid, affordable, and straightforward, which provided potential applications in pharmaceutical research and plant taxonomy.