Characterization of Organic Compounds as Complex Mixtures in the Environment: Method Development and Applications

数十万种有机化学品已被生产并推向全球市场。大量可能释放到环境中的人造化学品及其转化产物使人们很难表征这种复杂的有机混合物的环境风险。这是因为混合物中的大多数成分都是未知的，或者环境化学家和风险评估员不太了解。在环境中鲜为人知的化学物质中，有些可能对生态系统和人类人口构成高风险。有必要确定高风险因素，并说明其来源、环境命运和影响。我的研究计划的长期目标是确定在环境中具有高潜在风险的鲜为人知的有机污染物，并描述它们的来源、环境命运和影响。为了实现这一目标，我将使用基于高分辨率质谱学的技术进行非定向化学分析，并对有机化学品的环境命运和风险进行计算机模型模拟。将开发不同的方法，将从该仪器获得的信息与化学特性和环境危害联系起来。作为一个短期目标，将开发一个基于卷积神经网络的机器学习模型，将从非目标分析中获得的信息(GC和LC保留指数、准确质量、质量缺陷、同位素丰度、碎片损失)作为输入。模型的输出包括环境持久性、生物蓄积潜力和毒性。卷积神经网络将用经验数据和电子预测数据进行训练。另一种待开发的方法是将非靶标分析中检测到的所有可能的化学峰结构投影到可以揭示所检测化学品的环境危害的化学空间。使用这些方法，可以根据对环境的危害对检测到的化学品进行排序和选择。此外，还将根据环境因素随时间、空间和其他趋势的系统变化来确定不太为人所知的环境问题化学品。为从非目标分析中提取重要化学品而开发的不同方法将适用于来源区域的大气和水样。所开发的方法揭示的高潜在风险的化学品将成为进一步评估其来源、环境命运和影响的候选对象。围绕着用新开发的方法识别潜在环境问题的化学品的目标，拟议的研究计划将促进将化学应激源描述为复杂混合物对生态系统的影响，并在化学品达到可能不可逆转地恶化生态系统和影响人类健康的阈值之前提供早期预警。本研究还将为化学物质管理提供科学依据。

Hundreds-of-thousands of organic chemicals have been produced and introduced to global market. The large number of man-made chemicals potentially released to the environment together with their transformation products makes it difficult to characterize the environmental risks from the complex organic mixture. This is because most constituents of the mixture are unidentified or lesser-known to environmental chemist and risk assessors. Among the lesser-known chemicals in the environment, some can possibly pose high risk to the ecosystem and human population. There is a need to identify the high-risk ones and characterize their sources, environmental fate and impact. The long-term objective of my research program is to identify lesser-known organic contaminants with high potential risk in the environment and to characterize their sources, environmental fate and impact. To achieve the objective, I will use high-resolution mass spectrometry based techniques for non-targeted chemical analysis and computer model simulations on environmental fate and risk of organic chemicals. Different methods will be developed to link information gained from the instrument with chemical properties and environmental hazards. As one short-term objective, a convolutional neutral network based machine-learning model will be developed with information gained from non-targeted analysis as input (GC and LC retention indices, accurate mass, mass defect, abundances of isotopes, fragment losses). Output of the model includes environmental persistence, bioaccumulation potential and toxicity. The convolutional neutral network will be trained with empirical and in silico predicted data. Another method to be developed is to project all possible structures of chemical peaks detected in non-targeted analysis onto chemical space that can disclose environmental hazards of detected chemicals. Using the approaches, detected chemicals can be ranked and selected based on their environmental hazards. In addition, lesser-known chemicals of environmental concerns will also be identified based on their systematic variation with temporal, spatial and other trends in environmental factors. The different approaches developed for retrieval important chemicals from non-targeted analysis will be applied to atmospheric and aquatic samples in source regions. Chemicals with high potential risk revealed by the developed methods will become candidates for further assessment on their sources, environmental fate and impact. Revolving around the objective of identifying chemicals of potential environmental concerns with newly developed methods, the proposed research program will facilitate the characterization of the influence of chemical stressors as complex mixtures on the ecosystems and provide an early warning before chemicals reaching thresholds that would irreversibly deteriorate ecosystem and influence human health. This research will also provide scientific support for chemical management.