Project 4: Evaluating mediation effects of the microbiome and epigenetics using high dimensional assays

项目4摘要 高维度的人类微生物组和DNA甲基化数据为促进人类基因组的研究提供了巨大的希望。 了解无数人类疾病的潜在病因。中介建模是一个关键工具 在分子流行病学中用于推断生物过程的因果途径。到目前为止， 尚未广泛应用于微生物组和表观基因组的研究，尽管它可能会澄清 它们在疾病发病机制中的关键作用。据我们所知，没有可用的中介模型可供测试 人类微生物组是否介导疾病的发生。出现了使用中介建模的障碍 从微生物组的组成、遗传层次、稀疏和高维结构 数据另一个复杂程度是，介导可以通过单个微生物的变化发生， 通过改变微生物组的整体群落结构。对于DNA甲基化数据，模型 存在分析中介效应;然而，目前的方法依赖于参考数据来调整细胞- 成分效应。然而，参考数据往往无法获得，而且获取成本很高。无参考 已经提出了关联分析的方法来解决这个问题，但是这些方法还没有被 用于调解分析。为了应对这些重大挑战，我们将开发新的调解方法， 分析人类微生物组和DNA甲基化作为复杂介质的高维数据， 致病途径。我们将应用我们的模型来测试婴儿肠道微生物组，乳腺癌， 牛奶微生物组、脐带血DNA甲基化组和母乳DNA甲基化组在介导相关性中的作用 产前暴露（如砷暴露）与第一年儿童感染和过敏/特应性之间的关系 利用正在进行的大型纵向分子流行病学新罕布什尔州出生 队列研究。将开发R包来实现这两个模型。这些方法将使 确定疾病途径的复杂介质，以突出设计干预措施的机会， 支持儿童的健康和发展。

PROJECT 4 ABSTRACT High dimensional human microbiome and DNA methylation data offer great promise to contributing to the understanding of the underlying etiology of a myriad of human diseases. Mediation modeling is a critical tool used in molecular epidemiology to infer causal pathways for biological processes. As yet, mediation modeling has not been extensively applied to studies of the microbiome and epigenome even though it is likely to clarify their critical roles in disease pathogenesis. To our knowledge, there are no available mediation models to test whether the human microbiome mediates disease occurrence. Impediments to using mediation modeling arise from the compositional, phylogenetically hierarchical, sparse, and high dimensional structure of microbiome data. Another level of complexity is that mediations can occur through changes in individual microbes or through alterations to the overall community structure of the microbiome. For DNA methylation data, models exist for analyzing mediational effects; however, current methods rely on reference data to adjust for cell- composition effects. Yet reference data are often not available and are costly to obtain. Reference-free approaches have been proposed for association analyses to resolve this issue, but these have not been applied to mediation analyses. To address these critical challenges, we will develop new mediation methods to analyze high-dimensional data on the human microbiome and DNA methylation as complex mediators in disease causing pathways. We will apply our models to test the effects of the infant gut microbiome, breast milk microbiome, cord blood DNA methylome, and breast milk DNA methylome in mediating the associations between prenatal exposures (e.g. arsenic exposure) and childhood infections and allergy/atopy in the first year of life using the rich data from the large ongoing longitudinal molecular epidemiologic New Hampshire Birth Cohort Study. R packages will be developed to implement these two models. These methods will enable the identification of complex mediators of disease pathways to highlight opportunities for designing interventions to support children's health and development.