Methods for Gene-Enviroment Interactions Involving Gene Expression

描述(申请人提供)：尽管识别与疾病特征相关的基因变异的努力已经导致了相当数量的新发现的疾病SNPs，但在这一领域仍有许多开放的挑战。这些GWASNP倾向于解释性状中的低比例变异，并且在功能上具有模棱两可的作用。一些出版物提供的证据表明，同时考虑相关组织类型的遗传变异和基因表达有助于更全面地描述复杂人类特征的分子基础。特别是，基因表达变异同时受到遗传和环境的影响，这两个因素最终都会导致复杂的特征。因此，一个非常有意义的问题是能够定量地表征遗传、环境及其对基因表达变异的交互作用。我们建议建立一个量化框架，用于发现和剖析基因与环境(G×E)的交互作用，以解释高维特征的变异，如基因表达。我们还建议将该方法应用于几个前沿数据集，并提供软件，以便这些方法可以在未来的研究中得到广泛应用。我们要解决的关键挑战之一是解决相互作用的统计定义与生物定义不一致的问题。我们表明，统计定义并不是不变的，在尺度上的特征被放置。相反，我们发展了一个统计学定义，其中相互作用的存在与规模的变化无关，并且也与生物学定义一致。这对于基因表达数据尤其重要，在基因表达数据中，数据的分析规模是由技术决定的，而不是对复杂特征表现过程的直接物理衡量。

DESCRIPTION (provided by applicant): Although efforts to identify genetic variations associated with disease traits have lead to a respectable number of newly discovered disease SNPs, there are many open challenges in this area. These GWAS SNPs have tended to explain a low proportion of variation in the traits and have functionally ambiguous roles. A number of publications have provided evidence that simultaneously considering genetic variation and gene expression in relevant tissue types leads to a more comprehensive characterization of the molecular basis of complex human traits. In particular, gene expression variation captures influences from both genetics and environment, which both ultimately contribute to complex traits. Therefore, a problem of much interest is to be able to quantitatively characterize the genetic, environmental, and their interactive contributions to gene expression variation. We propose to build a quantitative framework for discovering and dissecting gene-by-environment (G x E) interactions explaining variation of high-dimensional traits, such as gene expression. We further propose to apply the methodology to several cutting edge data sets and also make software available so that the methods may be widely utilized in future studies. One of the key challenges that we will tackle is to resolve the inconsistency of the statistical definition of interaction with the biological definition. We show that the statistical definition is not invariant to changes in the scale on which the trait is placed. Instead we develop a statistical definition where the interaction exists irrespective of changes to the scale and also agrees with the biological definition. This is particularly important for gene expression data, where the scale on which the data are analyzed is determined by technology and not a direct physical measure of the process by which the complex trait is manifested.