Interpretable machine learning biodegradation of Direct Black 22 by Lysinibacillus sphaericus: Kinetics, mechanism, and wastewater application

초록

Bioremediation is a low-cost, environmentally friendly process that uses microorganisms to degrade harmful toxins, such as those from textile dyes. The current study investigated the efficiency of the isolated bacterium (IJA-2A), identified by 16S rRNA gene sequencing asLysinibacillus sphaericus, in remediating the Direct Black 22 (DB-22) dye. IJA-2A achieved 96% decolorization under optimal conditions (initial dye concentration of 100 mg L-1, pH 7.0, and 30 degrees C) within 96 h. Furthermore, increasing the dye concentration (500 mg L-1) resulted in a reduction in decolorization of only 30%. Machine learning models were trained to predict decolorization efficiency using regression. Among the machine learning models, the Support Vector Regression (SVR) model achieved a good mean R & sup2; of 0.876 (+/- 0.060) across 5-fold stratified cross-validation after hyperparameter optimization, with an RMSE of 7.61 and an MAE of 4.66. The SHAP values indicated incubation time as the most influential factor in predicting the decolorization efficiency of DB-22 dye. Furthermore, FT-IR spectroscopic studies demonstrated that molecular structural changes occurred during biodegradation. GC-MS analyses also provided evidence of intermediates produced throughout biodegradation, including sulfonated naphthalene, benzene derivatives, phthalates, and aniline. Lastly, this strain effectively removed color from actual textile effluent, achieving > 95% decolorization and a significant decrease in biological oxygen demand, chemical oxygen demand, and total organic carbon. A condition-adjusted comparison further indicated thatL. sphaericusIJA-2A achieved the highest predicted decolorization efficiency among 11 microbial species examined, highlighting its potential as a competitive candidate for Direct Black 22 bioremediation.

키워드