Data mining-based screening of prevalent mixture systems in aquatic environments: A case study of antibiotics in the Yangtze River Basin

Chemical pollution in real-world environment often involves exposure to combinations of thousands of chemicals. However, due to the vast number of possible combinations, it is nearly impossible to conduct comprehensive mixture toxicity tests and risk assessments for all of them. This study applied frequent itemset mining, a technique traditionally used in market basket analysis, to develop a prevalent mixture system screening (PMSS) method for identifying combinations that frequently co-occur in the environment. PMSS enables efficient data mining of chemical concentrations, allowing for the identification of a small number of prevalent mixture systems from numerous theoretical possibilities. In this study, 16 antibiotics were detected in the Linjiang River and the Xuebu River. Using the PMSS method, 48 prevalent antibiotic combinations (PACs), primarily ranging from binary to septenary combinations, were identified in the Xuebu River and the Linjiang River. The PACs in the surface water presented acceptable ecological risks, whereas the PACs in the sediments exhibited moderate to even high ecological risks. Therefore, targeted risk management measures should be developed for the sediments to reduce the potential harm to benthic organisms. Additionally, a case study demonstrates the application of identified PACs in mixture design. This study provides essential methodological and material support for advancing research on mixture toxicity evaluation and risk assessment.