Screening for frequently detected quaternary ammonium mixture systems in waters based on frequent itemset mining and prediction of their toxicity

• Published in: Ecotoxicology and environmental safety 2024-07, Vol.280, p.116581, Article 116581
• Main Authors: Tao, Meng-Ting, Sun, Xiao, Ding, Ting-Ting, Xu, Ya-Qian, Liu, Shu-Shen
• Format: Article
• Language: English
• Subjects: Combined toxicity; Data Mining; Direct equipartition ray design method; Disinfectants; Environmental Monitoring - methods; Microplate toxicity analysis; Mode of action; Quaternary Ammonium Compounds - toxicity; Toxicity Tests - methods; Uniform design ray method; Water Pollutants, Chemical - toxicity
• ISSN: 0147-6513; 1090-2414; 1090-2414
• DOI: 10.1016/j.ecoenv.2024.116581

Screening and prioritizing research on frequently detected mixture systems in the environment is of great significance, as conducting toxicity testing on all mixtures is impractical. Therefore, the frequent itemset mining (FIM) was introduced and applied in this paper to identify variables that commonly co-occur in a dataset. Based on the dataset of the quaternary ammonium compounds (QACs) in the water environment, the four frequent QAC mixture systems with detection rate ≥ 35 % were found, including [BDMM]+Cl−-[BTMM]+Cl− (M1), [BDMM]+Cl−-[BHMM]+Cl− (M2), [BTMM]+Cl− -[BHMM]+Cl− (M3), and [BDMM]+Cl−-[BTMM]+Cl−-[BHMM]+Cl− (M4). [BDMM]+Cl−, [BTMM]+Cl−, and [BHMM]+Cl− are benzyl dodecyl dimethyl ammonium chloride, benzyl tetradecyl dimethyl ammonium chloride, and benzyl hexadecyl dimethyl ammonium chloride, respectively. Then, the toxicity of the representative mixture rays and components for the four frequently detected mixture systems was tested using Vibrio qinghaiensis sp.-Q67 (Q67) as a luminescent indicator organism at 0.25 and 12 h. The toxicity of the mixtures was predicted using concentration addition (CA) and independent action (IA) models. It was shown that both the components and the representative mixture rays for the four frequently detected mixture systems exhibited obvious acute and chronic toxicity to Q67, and their median effective concentrations (EC50) were below 7 mg/L. Both CA and IA models predicted the toxicity of the four mixture systems well. However, the CA model had a better predictive ability for the toxicity of the M3 and M4 mixtures than IA at 12 h. •Frequent itemset mining was introduced to screen environmental detection data.•Four frequently detected QAC mixture systems were found.•Components of the four mixture systems showed obvious acute and chronic toxicity to Q67.•Toxicity of the four mixture systems was close to their components, with EC50 below 6 mg/L.•CA and IA models could predict the toxicity of the four frequently detected mixture systems.