Based on T.E.S.T toxicity prediction and machine learning to forecast toxicity dynamics in the photocatalytic degradation of tetracycline

Abstract

The integration of photocatalysis and biological treatment provides an effective strategy for controlling antibiotic contamination, which requires precise monitoring of toxicity changes during the photocatalytic process. In this study, nanoscale TiO₂ (P25) was employed to degrade tetracycline (TC) under full-spectrum irradiation, with O₂ identified as a crucial reactant for the generation reactive oxygen species (ROS). The toxicity simulation results of the degradation intermediates were closely correlated with the predictions of T.E.S.T software. By analyzing the content of intermediates under different experimental conditions, we developed a machine learning model utilizing the random forest algorithm with a correlation coefficient of R² = 0.878 and a mean absolute error of MAE = 0.02. The model can track the changes of photocatalytic intermediates, in combination with toxicity simulation, which facilitates the prediction of toxicity at different degradation stages, thus allowing selection of the optimal timing of biological treatment interventions.