Occurrence and removal prediction of pharmaceuticals positively correlated with antibiotic resistance genes in wastewater treatment processes

Abstract

Wastewater treatment plants (WWTPs) are important source of pharmaceuticals in the environment, identifying their relationship with antibiotic resistance genes (ARGs) and removal behavior is important for risk control. In this study, the correlations of common pharmaceuticals (including antibiotics and non-antibiotics) and ARGs both in wastewater and sludge were investigated, and machine learning was applied to predict the removal efficiencies of typical pharmaceuticals positively correlated with ARGs. Among the 23 kinds of target pharmaceuticals, both antibiotics and non-antibiotics were found to be positive with typical ARGs in wastewater (e.g. correlation coefficients between caffeine and blaTEM, aadD and qnrS were 0.738, 0.609, 0.936, p < 0.01) but negative in sludge (e.g. correlation coefficients between ofloxacin and tetO, tetW and qnrS were -0.922, -0.933 and -0.902, p<0.01), indicating the high risk of pharmaceuticals in wastewater promoting ARG spread. Furthermore, based on pharmaceutical removal efficiencies in different wastewater treatment processes in this and previous studies, the removal efficiencies of typical pharmaceuticals highly correlated with ARGs were well predicted through process operation parameters and wastewater characteristics by Random Forest model. HRT and temperature were identified as the most important explanatory variables. This study could provide comprehensive references for controlling pharmaceuticals in WWTPs.