Selectively Quantify\nToxic Pollutants in Water by\nMachine Learning Empowered Electrochemical Biosensors

Electroactive biofilm (EAB) sensors have become pivotal\nin water\nquality detection and early ecological risk warnings due to their\nremarkable sensitivity. However, it is challenging to identify multiple\ntoxicants in complex water bodies concurrently. This research developed\nan innovative biosensor detection strategy combined with machine learning.\nTo simultaneously quantify and qualitatively predict the presence\nof each toxin in a multitoxic system, we developed a prediction model\n(MEA-ANN) based on machine learning analysis of EABs by analyzing\nthe electrochemical toxicity response parameters of various toxicants\n(Cd<sup>2+</sup>, Cr<sup>6+</sup>, triclosan, and trichloroacetic\nacid). Furthermore, the mean impact value was utilized to filter the\ncharacteristic response parameters of toxicants, enhancing the prediction\naccuracy and efficiency of the model. The optimized model (OMEA-ANN)\ndemonstrated strong performance in predicting target toxicants within\ninterference systems containing analogs. The practicability and feasibility\nof this model were validated using seven real water samples and spiked\nnatural water samples, achieving R<sup>2</sup> > 0.9. The novel,\neco-friendly,\nand intelligent water ecological risk early warning strategy presented\nin this paper addresses the limitations of traditional EAB sensors.\nIt expands the applicability of EAB sensors for detecting multiple\ntoxicants in water, significantly advancing their role in water quality\nmonitoring. This approach provides valuable insights for the intelligent\nmanagement of sewage.