Bioremediation of toxic selenium from aqueous solution using Bacillus selenatarsenatis 9470T and machine learning approach

Abstract

The study focuses on a comprehensive study of the removal of selenium from an aqueous solution using Bacillus selenatarsenatis 9470^T and developing a predictive model under laboratory conditions. The growth rate and removal efficiency of the bacterium were investigated by varying pH, incubation time, inoculum dosages, and temperature. The characterization of Bacillus selenatarsenatis 9470^T before and after selenium exposure was performed using SEM, EDX, FT-IR, pH_zpc, and BET surface area. At 37 °C, the bacterium removed ∼99 % of selenium from the broth with initial concentrations below 10 Se-mg/L at pH 6 and 4 % inoculum dose within 20 h. The pH_zpc of Bacillus selenatarsenatis 9470^T was 7, and the specific surface area, average pore diameter, and pore volume were found to be 4.309 m²/g, 7.1 nm, and 7.6×10⁻⁹ m³/g, respectively, indicating the presence of mesopores. Langmuir isotherm and pseudo-second-order fit well, indicating monolayer sorption and chemisorption, respectively. The Regression Learner tool shows the best fit for Gaussian Process Regression (GPR), followed by Neural Network (NN), Support Vector Machines (SVM), and Tree (T). The best-fitted model (GPR) used for predicting unknown data points gave similar results to those given by the experiment. The study with competing ions affected Se removal efficiency as follows: manganese > iron > zinc > aluminum. The investigation showed that 89.7 % and 85 % of selenium were removed from sump 1 and sump 2, respectively. Moreover, it showed the ability to neutralize and remove other pollutants such as Fe, Mn, Zn, and Mg from real mine water samples. The experiment showed that this bacterium could be utilized to develop a sustainable approach for treating selenium-contaminated wastewater.