Machine Learning Accelerating the Condition Screening of Ceftriaxone Sodium Anaerobic Co-Metabolic Degradation

Anaerobic co-metabolism is a biotechnological process that improves biodegradation efficiency of refractory organics. By adding cosubstrates, it provides additional carbon sources and energy for the microbial metabolic degradation of organic matter. However, a large number of repeated experiments are required to screen out suitable conditions. It is typically lengthy and carries significant uncertainty. In this study, machine learning (ML) was used to drive the screening of anaerobic degradation conditions for ceftriaxone sodium (CTX) in wastewater treatment. The results showed that the XGBoost algorithm was able to effectively predict the decomposition efficiency with an accuracy of up to 95%. A Shapley additive explanation (SHAP) analysis showed that temperature, pH, and CTX/glucose ratio had the greatest impacts on the removal efficiency of CTX, thus highlighting the remarkable ability of ML to accelerate the screening of optimal degradation conditions. A high-throughput analysis proved that the dominant genera and microbial structures presented under the two environmental conditions with the largest significant difference (temperature, CTX/glucose ratio) were significantly different and that bacterial genera such as Fastidiosipila, norank_f_Prolixibacteraceae, norank_f_Bacteroidetes_vadinHA17, and Georgenia significantly affected the hydrolysis and acidification process. This work stands out by integrating advanced ML techniques into environmental engineering, thereby enhancing efficiency and providing richer analytical insights compared to traditional methods.