中文摘要：

化学品污染对人类健康和生态环境造成潜在风险。但是,危害性信息缺失是进行化学品风险评价的主要挑战。经济合作与发展组织（OECD）和美国环保署都提倡用非动物实验替代方法来弥补数据缺失。定量结构-活性关系（QSAR）被认为是一种有应用前景的替代技术。水生生物急性毒性是化学品风险评估和优先污染物筛选中最常用的参数之一。但是,目前可获得的实验毒性数据非常有限。本文总结了近年来发展的急性毒性预测模型,包括：（1）基于同类化合物建模;（2）基于数理统计建模;（3）基于化合物毒性作用模式建模。从模型预测能力、应用域、机理解释等角度对这3类模型进行了比较。其中,基于作用模式构建的模型一般具有较好的预测性能,并有助于机理解释,将是今后水生生物急性毒性预测的发展方向。

英文摘要：

Chemical contaminations lead potential risks to both human health and ecological environment. However, the lack of available data on the hazardous properties of chemicals is the major challenge for the risk assessment of chemicals. Non-animal alternative methods are encouraged to fill in data gaps by OECD and US EPA. Quantitative structure-activity relationship （QSAR） approach is regarded as one promising alternative technique. Information on acute toxicity to aquatic organisms are commonly used in the risk assessment and screening of priority substances. But, the available experimental toxicity data are very limited currently. In this paper, three types of prediction models of acute toxicity are summarized, including （1） models for particular chemical classes; （2） statistically derived models that are developed without an a priority mechanistic hypothe- sis; （3） models for a given mode or mechanism of action （MOA）. The predictive ability, applicability domain and mechanism interpretation of the three type models are compared. QSAR models based on MOA, which generally demonstrate rather good predictive ability and facilitate the interpretation of mechanism meanwhile.will become the main trend for predicting acute toxicity to aquatic organisms.