Machine Learning-Driven Electrochemical Aptasensing Platform for Highly Accurate Prediction of Phthalate Concentration in Multiple River Sites

Abstract

DEHP (di(2-ethylhexyl) phthalate), a widely used plasticizer, contaminates water through plastic waste leaching, posing severe health risks including growth delays and cardiovascular disease. Herein, we employed electrochemical aptasensors to analyze DEHP concentrations at the upper, mid, and lower layers of 3 sites across South Korean rivers. However, the solely sensor application faced challenges to classify and predict DEHP due to signal drift, biofouling, and limited specificity, especially with pH variations. Given these concerns, a machine learning (ML)-powered approach was applied, including a Conventional Generative Adversarial Network (cGAN) model for data augmentation and a hybrid Phthalate Boosting (PLBoost) algorithm for a robust multi-layer concentration analysis. The ML-powered electrochemical aptasensing platform significantly improved the DEHP prediction accuracy (97.11%) compared to those of the Liquid-liquid extraction/gas chromatography/mass spectrometry (LLE-GC-MS) measurement, minimizing the fluctuating conditions. Thus, an integration of the PLBoost with electrochemical aptasensors provides a robust DEHP monitoring platform in water samples.