Statistical methods for integrative analysis of multiple microbiome datasets

项目摘要： 最近的研究强调了人类相关微生物群在许多疾病和健康中的重要性 条件然而，在许多领域，由于样本量有限， 异质研究群体（例如，不同的种族、性别、年龄），以及技术可变性（例如，实验，试验 谈话/分析流水线）。例如，在艾滋病研究中，越来越多的证据表明肠道生态失调 会导致艾滋病相关的复发。然而，对其特征仍缺乏共识， 例如HIV感染是否增加或减少肠道中的微生物生物多样性以及哪些类群不同 HIV+和HIV-之间的区别综合分析，它汇总了来自多项研究的信息， 样本量和提升能力，是必要的，以推动该领域朝着一致和可重复的方向发展。 有可能提出预防和治疗干预的封面。然而，这对 由于研究之间的差异偏倚和测量误差而带来的统计挑战。 本提案的目的是开发和验证用于综合分析多种 微生物组数据集可能使用不同的实验室和预处理程序生成。我们 将使用研究特定的特征，例如研究人群、实验室和预处理管道， 并开发新的统计模型，用于表征微生物α（样本内）多样性，β （样本间）多样性和丰度（目标1）。我们将从微生物组质量分析数据 控制项目，一个大型的社区工作，通过多个管道对同一组样本进行测序， 旨在确定影响微生物组测序数据的技术变量，并以此为基础， 确定如何最好地利用所提出的方法中的信息（目标2）。 我们将把所提出的方法应用于HIV微生物组再分析项目，在该项目中，我们汇编了所有 我们将利用现有的HIV肠道微生物组16 s rRNA基因测序数据进行全面评估。我们将 我们还将我们提出的方法应用于从环境研究中的多个队列收集的微生物组数据。 在儿童健康结果的影响（ECHO）调查微生物组在影响健康的作用， 儿童和青少年。我们预计，所提出的方法将对几乎所有领域产生广泛的影响， 微生物组研究，并为分析16 s rRNA测序数据提供基础。

Project Abstract: Recent research has highlighted the importance of human associated microbiota in many diseases and health conditions. However, in many areas, results are often inconsistent across studies due limited sample sizes, heterogeneous study populations (e.g., different race, gender, age), and technical variability (e.g., experimen- tal/analysis pipelines). For example, in HIV studies there is increasing evidence suggesting that gut dysbiosis contributes to HIV-associated inflammation. However, there is still a lack of consensus on its characteristics, such as whether HIV infection increases or decreases the microbial biodiversity in the gut and which taxa differ between HIV+ and HIV-. Integrative analysis, which aggregates information from multiple studies to increase the sample sizes and boost power, is necessary to move the field forward toward consistent and reproducible dis- coveries with the potential of suggesting prophylactic and therapeutic intervention. This, however, poses serious statistical challenges due to the differential biases and measurement error between studies. The objective of this proposal is to develop and validate statistical methods for integrative analysis of multiple microbiome datasets that are potentially generated using different laboratory and pre-processing procedures. We will use the study-specific characteristics, such as study populations, laboratory and pre-processing pipelines, and develop novel statistical models for characterizing changes in microbial alpha (within-sample) diversity, beta (between-sample) diversity, and abundances (Aim 1). We will analyze the data from the microbiome quality control project, a large community effort that sequenced the same set of samples through multiple pipelines, designed to identify technical variables that impact the microbiome sequencing data, and use this as a basis to determine how to best use the information in the proposed methods (Aim 2). We will apply the proposed methods to the HIV microbiome re-analysis project, in which we have compiled all available 16s rRNA gene sequencing data for gut microbiome in HIV for a comprehensive evaluation. We will also apply our proposed methods to the microbiome data collected from multiple cohorts from the Environmental influences of child health outcomes (ECHO) to investigate the role of microbiome in impacting the health of children and adolescents. We expect that the proposed methods will have broad impact on almost all areas of microbiome research and provide a foundation for analyzing 16s rRNA sequencing data.