Integration of Omic Data to Estimate Mediation or Latent Structures

项目2：整合OMIC数据以估计中介或潜在结构 摘要 组学时代即将到来，基于人口的研究正在迅速衡量多种类型的数据， 探索风险因素和结果之间的潜在联系。从补充数据中整合数据 使用新的统计方法的研究途径将导致每个研究领域内的发现， 探索之间的区域，并推动创新。总体而言，该项目侧重于开发 用于整合先验怀疑对疾病起作用的多个组学数据的统计方法 或特质结果通过中介或潜在的结构化模型。这些方法涵盖了对研究的分析， 对相同个体的多个omic测量，以从多个omic测量的omic数据中汇总统计 问题研究在目标1中，我们将开发一个多组学因果推理测试（CIT），以促进其在大规模应用中的应用。 对个体进行测量的多组学数据集，以同时模拟多个风险因素和多个 调解员在目标2中，我们将开发一个综合模型来估计潜在的未知集群， 结合多种类型的OMIC测量，或者在横截面上测量，或者在多个时间点测量， 联合估计与关注结果相关的亚组。在目标3中，我们将估计联合因果效应 使用多个SNP和多个SNP的汇总统计量， 中间体的我们将在整个项目中利用其他项目的方法学发展 项目，并利用专业知识和援助，从计算和翻译的核心，我们将开发 强大的，计算效率高，用户友好的软件，适用于应用项目。总体而言，这些 方法将通过促进更好地理解潜在的 通过识别新的因素、估计联系来驱动潜在癌症病因的生物学机制 在这些因素之间，并确定具有潜在不同相关性的个体亚组 机制等

Project 2: Integration of Omic Data to Estimate Mediation or Latent Structures Abstract The omic era is upon us and population-based studies are moving rapidly to measure multiple types of data to explore the underlying connection between risk factors and outcomes. Integration of data from complementary avenues of research using novel statistical approaches will result in discoveries within each area of research, probe the area between, and push innovation forward. Overall, this project focuses on the development of statistical approaches for the integration of multiple omics data that are suspected, a priori, to act on a disease or trait outcome via mediation or a latent structured model. The approaches span the analysis of studies with multiple omic measures on the same individuals to summary statistics from omic data measured from multiple studies. In Aim 1, we will develop a multi-omic causal inference test (CIT) to facilitate its application to large multi-omic datasets measured on individuals to simultaneously model multiple risk factors and multiple mediators. In Aim 2, we will develop an integrative model to estimate latent unknown clusters aiming to incorporate multiple types of omic measures either measured cross-sectionally or at multiple time points to jointly estimating subgroups relevant to the outcome of interest. In Aim 3, we will estimate joint causal effects of intermediate factors or latent-outcome associations using summary statistics for multiple SNPs and multiple intermediates. We will leverage methodological developments from other projects within the overall program project and, using expertise and assistance from the computational and translation cores, we will develop robust, computationally efficient, and user-friendly software for application to applied projects. Overall, these methods will have a direct impact on applied investigations by facilitating a better understanding of potential biological mechanisms driving underlying cancer etiology via identifying novel factors, estimating connections between those factors, and identifying subgroups of individuals with potentially different associated mechanisms.