Statistical Models In Toxicology And Biochemistry

该项目的重点是：(1)开发将“尖端”研究成果纳入未来风险评估的方法；(2)开发改进风险评估的研究设计方法，特别是在涉及机械数据的情况下；(3)开发评估暴露、剂量-反应形状和效力的方法；(4)开发当涉及多种机制时评估混合物的方法；(5)开发协调癌症和非癌症健康风险评估的方法；(6)通过专家小组、同行审查和合作研究，直接让监管机构参与进来；(7)通过将理论发展与精确和方便的计算方法仔细联系起来，切实改进随机过程；(8)与NIEHS内的研究小组和从事类似工作的研究小组合作，以提高对疾病发病率的生物学理解；(9)通过假设检验和实验室研究，反复改进疾病发病率模型的生物学基础；(10)以科学可信的方式将疾病发病率模型与毒物动力学模型联系起来；(11)在模型的开发及其应用中使用最广泛的数据；(12)支持国家毒理学计划。 我们开发了一种定量的、统计上合理的方法，用于使用基因表达数据集分析可疑的基因调控网络。该方法基于贝叶斯网络，提供了一种直接量化基因表达网络和测试关于该网络中基因之间联系的假设的方法。进行了模拟研究，以评估该方法在小样本中的行为，并解决旨在量化基因相互作用网络的未来研究的设计。利用0.1、1.0和10.0 nM浓度的TCDD作用24小时后HPL1A肺呼吸道上皮细胞基因表达的变化，分析了一个假想的基因表达网络。该方法支持假设的网络，并允许评估将通常的二恶英表达变化与维甲酸受体系统联系起来的假设(参见下面的研究主题2.A)。 基因表达数据分析中尚未解决的主要问题之一是基因调控网络的识别。其他人已经提出了几种识别基因调控网络的方法，但这些方法侧重于使用多次配对比较来识别网络结构。我们开发了一种分析基因表达数据的方法，以确定与观察到的一组表达谱一致的调控结构。与其他方法不同，该方法通过使用基于似然的统计方法来获得与数据最一致的网络，从而超越了成对评估。贝叶斯方法(如上所述)然后可以用来量化基因之间的联系，以提供所产生的基因表达网络的完整特征。进行了仿真研究，以评估该方法的运行特性，并确定在不同设计策略下找到正确网络的概率。将该方法应用于小鼠胎儿纤维化MF129细胞中G(1)/S活化的数据。由此产生的基因-相互作用网络被用来识别节点基因并量化网络中基因之间的关系。使用对共同转录因子的搜索来验证所得到的网络。 我们还开发了计算机软件，使研究人员能够使用我们的贝叶斯网络分析方法来分析基因表达数据。使用了用户友好的、基于Windows的格式，使研究人员可以轻松地选择分析选项、定义网络结构和评估结果分析。

This Project focuses on (1) The development of methodology for incorporating "cutting-edge" research findings into future risk assessments; (2) The development of methods for designing studies to improve risk estimates, especially when mechanistic data is involved; (3) The development of methods for the evaluation of exposure, dose-response shape and potency; (4) The development of methods for evaluating mixtures when multiple mechanisms are involved; (5) The development of methods which harmonize cancer and non-cancer health risk assessments; (6) Direct engagement of the regulatory community through expert panels, peer review and collaborative research; (7) Practical improvement of stochastic processes through careful linkage of theoretical developments with computational methods that are accurate and convenient; (8) Collaboration with research groups within the NIEHS and research groups doing similar work to improve the biological understanding of disease incidence; (9) Iterative improvement of the biological basis for disease incidence models through a process of hypothesis testing and laboratory research; (10) Linkage of disease incidence models to toxicokinetics models in a scientifically credible manner; (11) Use of the broadest array of data in both the development of the model and its application; (12) Support of the National Toxicology Program. We developed a quantitative, statistically sound methodology for the analysis of suspected gene regulatory networks using gene expression data sets. The method is based on Bayesian networks and provides a means to directly quantify gene-expression networks and test hypotheses regarding the linkages between genes in this network. Simulation studies were performed to evaluate the behavior of this method for small samples and to address the design of future studies aimed at quantifying gene-interaction networks. Using gene expression changes in HPL1A lung airway epithelial cells after exposure to TCDD at levels of 0.1, 1.0 and 10.0 nM for 24 hours, a hypothesized gene expression network was analyzed. The method supports the assumed network and allowed the evaluation of a hypothesis linking the usual dioxin expression changes to the retinoic acid receptor system (see Research Theme 2.A below). One of the major unresolved issues in the analysis of gene expression data is the identification of gene regulatory networks. Several methods have been proposed by others for identifying gene regulatory networks, but these methods focus on the use of multiple pairwise comparisons to identify the network structure. We developed a method for analyzing gene expression data to determine a regulatory structure consistent with an observed set of expression profiles. Unlike other methods, this method goes beyond pairwise evaluations by using likelihood-based statistical methods to obtain the network that is most consistent with the data. Bayesian methods (as above) can then be used to quantify the linkages between genes to provide a complete characterization of the resulting gene-expression network. Simulation studies were performed to evaluate the operating characteristics of the method and to determine the probabilities of finding the correct network under different design strategies. This method was applied to data on G(1)/S activation in mouse fetal fibrosis MF129 cells. The resulting gene-interaction network was used to identify the nodal genes and quantify the relationships among genes within the network. Searches for common transcription factors were used to validate the resulting networks. We have also developed computer software to enable researchers to use our Bayesian networks analysis method to analyze gene expression data. A user-friendly, windows-based format was used to make it easy for researchers to choose options for the analysis, define network structures and evaluate the resulting analysis.