Mathematical prediction models for inactivation of antibiotic-resistant bacteria in kitchen wastewater by bimetallic bionanoparticles using machine learning with gene expression programming

Abstract

Listen

Translate article

The current study investigated the mathematical prediction model for the inactivation of antibiotic-resistant bacteria (B. subtilis, S. aureus, and E. coli) in kitchen wastewater by bimetallic bionanoparticles (Zn/Cu NPs) using gene expression programming (GEP). The interaction between Zn/Cu NP concentration, time, pH, initial cell concentration, and agitation speed was analysed based on the prediction tree. The best prediction model was obtained with GEP2 with root-mean-square error (RMSE) values of 0.2943 vs. 0.2988 for E. coli and 0.4059 vs. 0.4267 for S. aureus, and GEP4 for B. subtilis with RMSE value of 0.5123 vs. 0.4582 for the training and testing phase, respectively. The best log reductions were obtained with 15 μg mL−1 NPs for 60 min at pH 8 and 180 rpm; the predicted and actual log reductions of S. aureus, B. subtilis, and E. coli were 5.19 vs. 5.8, 5.73 vs. 5.67, and 6.00 vs. 5.99, respectively. These findings indicate that S. aureus exhibited more deviation in the predicted and actual results compared to B. subtilis and E. coli, which might be related to the behaviour of S. aureus towards the NPs based on their cell shape and their distribution in the real kitchen wastewater. The NP concentration and time are the main factors in the inactivation process. The NPs act on the bacterial cells externally by damaging the bacterial cell surface and transportation systems. Nps also act internally by altering the metabolic pathway in the bacterial cytoplasm. These findings confirm the effectiveness of GEP in predicting bacterial inactivation in kitchen wastewater.