Multivariate analysis of microbial absolute abundance in population-based studies.

项目总结 口腔微生物群已被证明与人类健康有关。大规模的流行病学研究， 包括我们正在进行的口腔感染、葡萄糖耐量和胰岛素抵抗研究(Origins)， 确定了与口腔疾病显著相关的口腔微生物区系。然而，与 口腔外疾病的风险，如全身炎症和糖尿病风险通常要小得多。什么 存在的信号通常通过多次测试校正或整体数据减少过程来衰减。 因此，由于缺乏统计能力，许多潜在的重要关联仍未被发现。 事实上，缺乏动力的问题在微生物组关联研究中很常见。缺乏系统性的 用于识别微生物失衡的统计方法，该微生物失衡涉及弱但潜在的细菌 对疾病的相加作用。在这项提案中，我们将开发新的统计方法并伴随 解决这一挑战的软件。我们将在以前工作的基础上，开发新的监督数据 约简方法，提高关联分析的能力。这些方法将首先利用临床 信息来选择临床上相关的分类群，然后聚集所选的分类群以形成少量的 微生物丰度得分。得分将用于进一步的关联分析，以减轻 多重测试校正，并增强来自单个细菌分类群的微弱信号。此外，这些方法 将直接适用于下一代测序读数的绝对丰度，避免了偏差 以及与相对丰度相关的功率损失。我们将对计数值的、零膨胀的 绝对丰度，利用测序数据中的丰富信息。新开发的方法将是 适用于原产地数据。我们希望确定与口腔微生物区系显著相关的 炎症和代谢生物标记物。这一潜在的发现有望提高我们对 疾病机制，促进炎症和代谢紊乱的早期诊断，并启发新的 治疗方法。在其他微生物组研究中的应用将有助于我们更好地理解 人类微生物群在疾病发展和/或维持健康中的作用。

PROJECT SUMMARY The oral microbiome has been shown to be related to human health. Large-scale epidemiological studies, including our ongoing Oral Infections, Glucose Intolerance and Insulin Resistance Study (ORIGINS), have identified oral microbiota that are significantly associated with oral diseases. However, the associations with extra-oral disease risks such as systemic inflammation and diabetes risk are usually much weaker. What signals exist are typically attenuated by multiple testing corrections or holistic data reduction procedures. Consequently, many potentially important associations remain undiscovered due to a lack of statistical power. In fact, the lack-of-power issue is common in microbiome association studies. There is a lack of systematic statistical approaches for identifying microbial imbalance that involves bacteria with weak but potentially additive effects on diseases. In this proposal, we will develop novel statistical methods and accompanying software to address the challenge. We will build upon our previous work and develop new supervised data reduction approaches to improve the power of association analysis. The methods will first leverage clinical information to select clinically relevant taxa, and then aggregate selected taxa to form a small number of microbiome abundance scores. The scores will be used for further association analysis to reduce the burden of multiple testing corrections and to enhance weak signals from individual bacterial taxa. Moreover, the methods will be directly applicable to absolute abundances of next-generation sequencing read counts, to avoid the bias and power loss associated with relative abundances. We will properly model the count-valued, zero-inflated absolute abundances to harness the rich information in sequencing data. The newly developed methods will be applied to the ORIGINS data. We expect to identify oral microbiota that are significantly associated with inflammatory and metabolic biomarkers. The potential discovery promises to improve our understanding of disease mechanisms, facilitate early diagnosis of inflammatory and metabolic disorders, and inspire new therapeutic approaches. The applications to other microbiome studies will improve our understanding of the role of human microbiome in the development of disease and/or maintenance of health.