HIERARCHICAL MODELING OF INTERACTIONS IN GENOME-WIDE AND PATHWAY-BASED STUDIES

描述（由申请人提供）：在全基因组和基于通路的关联研究中相互作用的分层建模：该资助的首要目标是调查在全基因组关联研究和基于通路的候选基因研究中基因-环境和基因-基因相互作用研究中分层建模的使用。我们将把我们的方法应用于两个大型NIH支持项目的数据：结肠CFR和儿童健康研究。来自这些关联研究的数据通常遵循基因区域内的多态性、子途径内的基因和病因学网络内的子途径的自然分层结构。通过建立一个统计模型来反映这种自然层次结构，我们的目标是更好地解释因素之间的依赖关系，并更好地结合我们对潜在病因的了解。在这个建议中，除了评估这些模型的统计形式和结构，我们还旨在衡量各种类型的先验信息和中间测量对推断的影响。对于全基因组关联研究，我们将开发分析方法，用于合并GxE相互作用，处理多重检验问题，并扩展到可能更有效的2阶段研究设计。这些方法也将扩展到测试环境因素与多个SNP的相互作用。特别是基于途径的研究，我们的目标是探索机制模型（如动力学模型）和分层回归模型的可行性和性能，以及子途径和跨途径网络的基因和环境因素的模型选择。我们使用本体或基于专家的关系数据库形式的先验知识来帮助制定数据分析的先验知识。此外，我们将研究各种多阶段采样方案及其与潜在基因组学数据的相互作用，包括全外显子表达，全基因组体细胞突变和潜在的生物标志物测量。最后，我们将比较我们的方法与各种数据挖掘技术，以允许基因在多个通路中起作用。总的来说，我们的目标是开发统计技术，使其能够检测哪些基因参与疾病，更重要的是，在哪些环境背景下，他们的行为。通过识别遗传和环境因素，我们将在理解导致疾病的潜在机制方面取得进展，并可能找到预防和治疗复杂疾病的方法。公共卫生相关性：总的来说，我们的目标是开发统计技术，正式纳入我们的生物学知识，并使其可行的检测哪些基因参与疾病，更重要的是，在哪些环境背景下，他们的行为。通过识别遗传和环境因素，我们将在理解导致疾病的潜在机制方面取得进展，并可能找到预防和治疗复杂疾病的方法。

DESCRIPTION (provided by applicant): Hierarchical Modeling of Interactions in Genome-Wide and Pathway-Based Association Studies: The overarching goal of this grant is to investigate the use of hierarchical modeling in the study of gene-environment and gene-gene interactions in both genome-wide association studies and pathway-based candidate gene studies. We will apply our methods to data available from two large NIH-supported projects: the Colon CFR and Children's Health Study. Data from these association studies often follows a natural hierarchical structure with polymorphisms within gene regions, genes within sub-pathways, and sub-pathways within etiologic networks. By building a statistical model to reflect this natural hierarchy we aim to better account for the dependencies between factors and to better incorporate our knowledge of the underlying etiology. In this proposal, in addition to evaluating the statistical form and structure of such models we also aim to gauge the impact of various types of prior information and intermediate measurements on inference. For genome-wide association studies we will develop analytic approaches for the incorporation of GxE interactions that deal with the multiple testing problem and extend to potentially more efficient 2-phase study designs. These methods will be expanded to test for the interaction of an environmental factor with multiple-SNPs, as well. Specifically with pathway-based studies, we aim to explore the feasibility and performance of mechanistic models (e.g. kinetic models) and hierarchical regression models with model selection for genes and environmental factors within sub-pathways and across networks of pathways. We use prior knowledge in the form of ontologies or expert-based relational databases to help formulate priors for the data analysis. Furthermore, we will investigate various multistage sampling schemes and their interplay with potential genomics data, including whole-exon expression, whole-genome somatic mutations and potential biomarker measures. Finally, we will compare our methods to various data mining techniques to allow genes to act within multiple pathways. Overall, we aim to develop statistical techniques that make it feasible to detect which genes involved in disease and, importantly, in which environmental context they act. By identifying both genetic and environmental factors, we will make progress in understanding the underlying mechanism that leads to disease and potentially identify ways in which to both prevent and treat complex diseases. PUBLIC HEALTH RELEVANCE: Overall, we aim to develop statistical techniques that formally incorporate our biologic knowledge and make it feasible to detect which genes are involved in disease and, importantly, in which environmental context they act. By identifying both genetic and environmental factors, we will make progress in understanding the underlying mechanism that leads to disease and potentially identify ways in which to both prevent and treat complex diseases.