Mechanism-Driven Virtual Adverse Outcome Pathway Modeling for Hepatotoxicity

项目摘要/摘要 评估肝毒性的实验动物和临床试验需要广泛的资源和 周转时间长。利用计算模型直接预测新化合物的毒性是一种 降低药物开发成本和筛选多种工业化学品的有希望的战略 以及目前缺乏安全评估的环境污染物。然而，目前计算的 复杂毒性终点的模型，如肝脏毒性，对于筛选新化合物是不可靠的 面临着无数的挑战。我们最近的研究表明，传统的定量结构-活性 关系建模适用于相对简单的属性或毒性端点，具有明确的 机制，但未能解决复杂的生物活性，如肝脏毒性。这样做的主要目标是 建议开发新的机制驱动的虚拟不良结局路径(VAOP)模型 以高通量的方式快速准确地评估肝毒性，由此产生的VAOP模型将 通过补充体外和体外测试进行实验验证。我们已经生成了一个 基于抗氧化反应元件(ARE)途径的VAOP模型的初步研究 使用体外测试进行初步验证和改进。为此，我们的项目将产生新的预测 通过将1)虚拟细胞应激途径模型应用于机制描述和 评估新化合物；2)计算预测以填补特定目标的缺失数据 途径内；3)三种互补生物检测的体外实验验证；以及4)体外实验 具有生化转化能力的原代人肝细胞池的实验验证。这个 这项研究的科学方法是开发一个通用的建模工作流，可以利用所有的 可用的短期测试信息，从使用新型机器的两个计算预测中获得 学习方法和体外实验，用于感兴趣的目标化合物。我们将验证并使用我们的 建模工作流，以直接评估新化合物的肝脏毒性并确定候选化合物的优先顺序 在原代人肝细胞池中的验证。由此产生的工作流程将通过门户网站进行传播 为世界各地的公众用户提供互联网接入。重要的是，这项研究将为下一步铺平道路 通过组合以下各项重新构建建模过程来生成化学毒性评估 大数据、计算建模和低成本的体外实验。据我们所知， 该项目的实施将导致第一个公开可用的机制驱动的建模和网络- 基于可公开访问的大数据的复杂化学毒性预测框架。这些 交付成果将对公众健康产生重大影响，不仅是为了安全测试或 新的化学发展，也揭示了毒性机制。

PROJECT SUMMARY/ABSTRACT Experimental animal and clinical testing to evaluate hepatotoxicity demands extensive resources and long turnaround times. Utilization of computational models to directly predict the toxicity of new compounds is a promising strategy to reduce the cost of drug development and to screen the multitude of industrial chemicals and environmental contaminants currently lacking safety assessments. However, the current computational models for complex toxicity endpoints, such as hepatotoxicity, are not reliable for screening new compounds and face numerous challenges. Our recent studies have shown that traditional Quantitative Structure-Activity Relationship modeling is applicable for relatively simple properties or toxicity endpoints with a clear mechanism, but fails to address complex bioactivities such as hepatotoxicity. The primary objective of this proposal is to develop novel mechanism-driven Virtual Adverse Outcome Pathway (vAOP) models for the fast and accurate assessment of hepatotoxicity in a high-throughput manner The resulting vAOP models will be experimentally validated using a complement of in vitro and ex vivo testing. We have generated a preliminary vAOP model based on the antioxidant response element (ARE) pathway that has undergone initial validation and refinement using in vitro testing. To this end, our project will generate novel predictive models for hepatotoxicity by applying 1) a virtual cellular stress pathway model to mechanism profiling and assessment of new compounds; 2) computational predictions to fill in the missing data for specific targets within the pathway; 3) in vitro experimental validation with three complementary bioassays; and 4) ex vivo experimental validation with pooled primary human hepatocytes capable of biochemical transformation. The scientific approach of this study is to develop a universal modeling workflow that can take advantage of all available short-term testing information, obtained from both computational predictions using novel machine learning approaches and in vitro experiments, for target compounds of interest. We will validate and use our modeling workflow to directly evaluate the hepatotoxicity of new compounds and prioritize candidates for validation in pooled primary human hepatocytes. The resulting workflow will be disseminated via a web portal for public users around the world with internet access. Importantly, this study will pave the way for the next generation of chemical toxicity assessment by reconstructing the modeling process through a combination of big data, computational modeling, and low cost in vitro experiments. To the best of our knowledge, the implementation of this project will lead to the first publicly available mechanisms-driven modeling and web- based prediction framework for complex chemical toxicity based on publicly-accessible big data. These deliverables will have a significant public health impact by not only prioritizing compounds for safety testing or new chemical development, but also revealing toxicity mechanisms.