Statistical Methods for Network-based Integrative Analysis of Microbiome Data

项目概要 过去十年在与人类相关的微生物组生物标志物的发现方面取得了实质性进展 人的健康和疾病。然而，尽管之前的工作令人兴奋，但我们目前仍然缺乏对 肠道微生物组影响人类健康的机制。一个突出的挑战是如何整合 微生物组和其他组学数据类型在微生物组多组学分析研究中生成，以阐明微生物组学 微生物功能途径。不幸的是，现有的综合分析统计方法并不能充分地 解决微生物组数据特有的分析挑战。微生物组数据是组合的、零膨胀的、 高维和高度结构化，其中样本通过生态定义的距离和分类群相关联 通过它们的系统发育相关。我们建议利用我们在网络分析和高维统计方面的专业知识 推理来解决微生物组数据分析特有的这些挑战。我们的总体目标是制定严格的 我们的统计方法可以产生将微生物功能途径与宿主相关的可靠而有力的推论 健康状况。所提出的方法是由我们与拉丁裔研究中心的合作推动的 狗衰老项目，并包括用于微生物和代谢组学联合分析的新颖推理程序 网络（目标 1），一种结合先验生物学知识的联合降维新方法 关于样本之间和变量之间的关系（目标 2），以及联合的强大框架 将微生物组和其他组学数据类型与健康结果相关联（目标 3）。我们将开发高效、 所提出方法的易于使用的软件工具（目标 4）。这项工作具有创新性和意义，因为它 将为微生物组的作用提供系统生物学见解，并有可能产生重大影响 新型微生物组生物标志物的鉴定。该提案的成功完成将产生重要的 tant 社区资源，包括综合分析的新方法及其用户友好的软件 工具。借助这些分析工具，该项目的长期目标是加快微生物组的发现 基于治疗目标，并有助于开发基于微生物组的精准疗法。

Project Summary The past decade has seen substantial progress in the discovery of microbiome biomarkers associated with hu- man health and diseases. However, despite the exciting prior work, we currently still lack an understanding of the mechanism by which the gut microbiome impacts human health. An outstanding challenge is how to integrate microbiome and other -omics data types generated in microbiome multi-omics profiling studies to elucidate mi- crobial functional pathways. Unfortunately, available statistical methods for integrative analysis do not adequately address the analytical challenges specific to microbiome data. Microbiome data are compositional, zero-inflated, high-dimensional, and highly structured where samples are related by ecologically defined distances and taxa are related by their phylogeny. We propose to use our expertise in network analysis and high-dimensional statistical inference to tackle these challenges unique to microbiome data analysis. Our overall objective is to develop rigor- ous statistical methods that yield reliable and powerful inferences relating microbial functional pathways with host health conditions. The proposed methodologies are motivated by our collaboration with the Study of Latinos and the Dog Aging Project, and include a novel inference procedure for joint analysis of microbial and metabolomic networks (Aim 1), a novel method for joint dimensionality reduction which incorporates prior biological knowledge about the relationships between samples and between variables (Aim 2), and a powerful framework for jointly associating microbiome and other -omics data types with health outcomes (Aim 3). We will develop efficient and easy-to-use software tools for the proposed methods (Aim 4). This work is innovative and significant, because it will provide systems biology insights into the role of the microbiome and has the potential to make a major impact on the identification of novel microbiome biomarkers. Successful completion of this proposal will generate impor- tant community resources, including new methodologies for integrative analysis and their user-friendly software tools. With these analytical tools, the longer-term goal of this project is to hasten the discovery of microbiome- based therapeutic targets and contribute to the development of microbiome-based precision therapies.