| 摘要 | 近年来，全球新污染物数量急剧增加，对水生生态系统造成了严重的危害，迫切需要开展新污染物水生态风险评估工作。由于新污染物在不同水质类型中的毒性存在较大差异，开展不同水质类型中的新污染物的风险评估有利于提高特定水质风险决策的准确性。然而，新污染物的毒性数据严重不足，仅通过实验方法无法满足新污染物的风险评估要求。因此，有必要构建生态风险评估相关参数的计算毒理学预测模型，综合考虑持久性和富集性等生物指标，发展基于综合评分的新型风险评估方法，为全面评估新污染物的生态风险提供有力支撑。
（1）本论文基于机器学习和特征筛选的方法，建立了8类海洋生物和8类淡水生物的定量-构效关系（quantitative structure-activity relationship, QSAR）毒性预测模型。16种水生生物模型具有良好的预测能力和稳健性，其训练集决定系数（RTR2）均达到了0.99以上，验证集决定系数（QEXT2）和留一法交叉验证系数（QLOO2）的平均值也分别达到了0.79和0.69，且有12类物种的QEXT2达到0.75以上，有6类物种的QLOO2达到0.80以上。机理解释发现，化合物的辛醇水分配系数（octanol-water partition coefficient, KOW）和氯原子的数量（number of chlorine atoms, nCl）是海水生物与淡水生物毒性差异的主要因素，并与盐度和生物体内的离子浓度有关。化合物的芳香性和结构复杂性是区分海洋生物中初级生产者和其他营养级的主要因素，主要是由于化合物与叶绿素的竞争性结合以及与不同酶间的相互作用有关。
（2）基于QSAR模型补充的数据，研究选取了两类水生生态系统典型新污染物全氟/多氟类化合物（per- and polyfluoroalkyl substances, PFASs）和有机磷酸酯（organophosphate esters, OPEs），探讨了两类污染物在海水与淡水中的毒性差异。OPEs和PFASs对海洋浮游动物的毒性高于浮游植物，OPEs和PFASs对海洋生物的毒性均高于淡水生物，且PFASs对海洋和淡水生物的毒性差异更大，其中六氟环氧丙烷四聚酸（hexafluoropropylene oxide tetramer acid, HFPO-TeA）高达3188倍。由此可见，未考虑毒性差异会导致以HFPO-TeA等为代表的PFASs类物质在海水中的风险被低估。研究还揭示了结构特征对OPEs和PFASs毒性的影响规律：芳环和氯等取代结构会增加OPEs的毒性，且烷基OPEs的毒性与其侧链长度呈正相关关系。PFASs的毒性与链长、取代基类型和醚键的数量密切相关。PFASs的毒性随链长的增加而增强，且含有磷酸基、磺酸基和两个醚键以上的PFASs毒性较高。
（3）本论文结合持久性、富集性、毒性和浓度四个指标，发展了基于综合评分的风险评估方法，评估了渤海地区13种PFASs和6种OPEs的生态风险。综合评分法结果与传统风险评估风险熵（risk quotients, RQs）法对比发现，全氟十二酸（perfluorododecanoic acid, PFDoDA）、氟（3,5-二氧六）酸（perfluoro(3,5-dioxahexanoic) acid, PFO2HxA）、磷酸三（2-氯丙基）酯（tris(2-chloropropyl)phosphate, TCPP）和磷酸三苯酯（triphenyl phosphate, TPP）排名得到了极大的提升，其中PFDoDA从最后一名上升为第三名。尽管基于RQs法的评估结果表明，部分OPEs（TCPP和TPP）和PFASs（PFDoDA和PFO2HxA）的风险较低，甚至无风险，但在渤海地区的实际影响可能更大。因此，基于综合评分法的风险评估可揭示某些化合物的潜在生态风险。 |
| 其他摘要 | In recent years, there has been a marked upsurge in the global emergence of emerging pollutants, inflicting considerable damage upon aquatic ecosystems. Addressing this pressing issue necessitates meticulous evaluations of the ecological perils associated with these emerging pollutants. The toxicity of emerging pollutants varies significantly across different water types, and conducting risk assessments for specific water types can improve the accuracy of risk-based decision-making. Nonetheless, the scarcity of toxicity data for emerging pollutants presents a significant challenge since relying solely on experimental methods is inadequate to meet the demands of risk assessments. It is therefore essential to develop computational toxicology models, considering biomarkers such as persistence and bioaccumulation, and developing a novel risk assessment method based on comprehensive scores to provide robust support for the comprehensive evaluation of the ecological risks posed by emerging pollutants.
(1) In this thesis, quantitative structure-activity relationship (QSAR) models for eight marine species and eight freshwater species were established based on machine learning and feature selection methods. The 16 models for aquatic species demonstrated exceptional predictive prowess and stability. The training set determination coefficients (RTR2) surpassed 0.99, while the external validation set determination coefficients (QEXT2) and leave-one-out cross-validation coefficients (QLOO2) had average values of 0.79 and 0.69, respectively. For 12 species, QEXT2 achieved a value of 0.75 or greater, while for six species, QLOO2 surpassed 0.8. Mechanistic analysis revealed that the octanol-water partition coefficients (KOW) and the number of chlorine atoms (nCl) were the primary factors contributing to toxicity differences between marine and freshwater species, which were related to salinity and ion concentration in organisms. Aromaticity and structural complexity of compounds were the main factors distinguishing primary producers from other trophic levels in marine organisms, mainly due to competitive binding of compounds to chlorophyll and interactions with different enzymes.
(2) Utilizing data from the QSAR models, this study focused on two categories of emerging pollutants commonly found in aquatic ecosystems: per- and polyfluoroalkyl substances (PFASs) and organophosphate esters (OPEs). It sought to investigate the variances in their toxicity levels between marine and freshwater environments. OPEs and PFASs were found to be more toxic to marine zooplankton than phytoplankton. Additionally, both OPEs and PFASs displayed higher toxicity in marine organisms compared to freshwater. Notably, PFASs showed an even more pronounced difference in toxicity between marine and freshwater species. For instance, the toxicity of hexafluoropropylene oxide dimer acid (HFPO-TeA) in marine species was 3188 times higher compared to that in freshwater species. Consequently, overlooking these toxicity differences may result in the underestimation of risks associated with PFASs. The study further uncovered the impact of structural characteristics on the toxicity of OPEs and PFASs. Aromatic rings and chlorine substituents enhanced the toxicity of OPEs, with the toxicity of alkyl OPEs showing a positive correlation to the length of the side chain. Meanwhile, the toxicity of PFASs was closely linked to chain length, the type of substituent, and the number of ether bonds. Heightened levels of toxicity were noted in PFASs characterized by elongated chains, phosphonate or sulfonate groups, as well as those possessing more than two ether bonds.
(3) In addition to the traditional risk assessment risk entropy (RQs) method, this thesis developed a comprehensive scoring-based risk assessment method by integrating four indicators of persistence, enrichment, toxicity and concentration to assess the ecological risk of 13 PFASs and 6 OPEs in the Bohai Sea. Compared with the RQs method, the composite scoring approach demonstrated a notable improvement in the ranking of perfluorododecanoic acid (PFDoDA), fluoro(3,5-dioxohexa)acid (PFO2HxA), tris(2-chloropropyl) phosphate (TCPP), and triphenyl phosphate (TPP), with PFDoDA moving up from last to third place. While the RQs method-based assessment results suggest that some OPEs (TCPP and TPP) and PFASs (PFDoDA and PFO2HxA) pose low or no risk, the actual impact in the Bohai may be more significant. Consequently, employing a risk assessment that utilizes an integrated scoring approach can effectively unveil the potential ecological risks associated with specific compounds. |
修改评论