A Machine Learning-based approach for Simultaneous Electrochemical Detection of Interfering Analytes

Abstract

Electrochemical biosensors can be used to detect analytes of importance precisely. These sensors generate rapid and accurate electrical signals that reveal the presence and concentration of the targeted analyte. Detecting multiple analytes simultaneously with an electrochemical biosensor is advantageous. It provides cost and time efficiency, multiplexing capability, and flexibility, making it valuable in diverse applications such as medical diagnostics, environmental monitoring, and industrial processes. However, simultaneous detection of analytes may suffer from the problem of interference. The interference effect causes the signal of an analyte at a particular concentration to deviate from the expected one. We observe a similar effect in the simultaneous detection of Folic Acid and Uric Acid using a nanomaterial-based electrochemical sensor. To address this effect, we propose a machine learning (ML) approach. ML algorithms handle complex interactions by autonomously identifying patterns, dependencies, and nonlinear relationships within data, enabling it to make predictions and decisions in intricate and dynamic scenarios. Our approach can be generalised to any two analytes showing interference and would scale well to interference between multiple analytes. We test several regression algorithms and compare their performance to the standard calibration plot method. As compared to the standard method, our approach shows a 4.49 {micro}M decrease in concentration prediction error.