Bayesian Machine Learning Tools for Analyzing Microbiome Dynamics

人类微生物区系在健康和疾病中发挥着重要作用，其治疗操作正在进行中 积极调查遍及美国国立卫生研究院所有研究所的各种疾病。我们的微生物区系天生就是 动态的，分析这些与时间相关的特性是将微生物区系与疾病强有力地联系起来的关键， 并预测针对微生物区系的治疗效果；事实上，纵向微生物组数据正在 收购的频率越来越高，是美国国立卫生研究院资助的许多项目的主要组成部分。然而，还有 目前缺乏用于分析微生物组时间序列数据的计算工具，这提供了几个 特殊挑战包括高测量噪声、不规则和稀疏的时间采样以及复杂的 变量之间的依赖关系。这项提议的目标是引入新的功能，改进 并提供用于分析动态或变化模式的工具的最新实现 微生物组时间序列数据。我们开发的工具将使用贝叶斯机器学习方法，即 因其强大的概念和实践优势而广为人知，特别是在生物医学领域。工具 将在人造和真实的人体微生物组数据上进行严格的测试和验证，包括公开 可用的数据集和来自合作者的提供16S rRNA测序、元基因组学和 代谢组学数据。我们提出了三个具体目标。对于目标1，我们将开发集成贝叶斯机器 预测微生物群的种群动态及其对扰动反应的学习工具。这些 工具将包括一种新的模型，它可以同时学习具有相似相互作用结构的微生物组 并预测它们随着时间的推移的行为，这包含了先前的系统发育信息。模型将是 通过将随机微生物动力学和整个测量过程中的误差纳入其中，进一步改进了 模特。对于目标2，我们将开发贝叶斯机器学习工具来根据微生物群预测宿主状态 动力学。这些工具将学习易于解释的、人类可读的规则，这些规则从 微生物组时间序列数据，并将进一步扩展，以处理各种纵向研究设计。为 目标3，我们将设计我们的微生物组动力学分析软件工具，以实现最佳性能、易用性 易用性、可维护性、可扩展性和对社区的传播。总的来说，拟议的工作将产生一个 一套用于分析微生物组动态的当代软件工具，预计将得到广泛使用和主要 冲击力。该软件将允许调查人员回答有关以下问题的重要科学和翻译问题 微生物组，包括发现哪些微生物分类群或它们的元基因组随着时间的推移受到 扰动，如饮食变化或病原体入侵；预测这些扰动的影响 随着时间的推移，包括肠道微生物区系的组成或稳定性的变化；以及在 预测人类宿主疾病风险的多组微生物组数据。

The human microbiota plays an important role in health and disease, and its therapeutic manipulation is being actively investigated for a wide range of diseases that span every NIH institute. Our microbiota are inherently dynamic, and analyzing these time-dependent properties is key to robustly linking the microbiota to disease, and predicting the effects of therapies targeting the microbiota; indeed, longitudinal microbiome data is being acquired with increasing frequency, and is a major component of many NIH-funded projects. However, there is currently a dearth of computational tools for analyzing microbiome time-series data, which presents several special challenges including high measurement noise, irregular and sparse temporal sampling, and complex dependencies between variables. The objective of this proposal is to introduce new capabilities, improve on, and provide state-of-the-art implementations of tools for analyzing dynamics, or patterns of change in microbiome time-series data. The tools we develop will use Bayesian machine learning methods, which are well-recognized for their strong conceptual and practical advantages, particularly in biomedical domains. Tools will be rigorously tested and validated on synthetic and real human microbiome data, including publicly available datasets and those from collaborators providing 16S rRNA sequencing, metagenomic, and metabolomics data. We propose three specific aims. For Aim 1, we will develop integrated Bayesian machine learning tools for predicting population dynamics of the microbiome and its responses to perturbations. These tools will include a new model that simultaneously learns groups of microbes with similar interaction structure and predicts their behavior over time, and that incorporates prior phylogenetic information. The model will be further improved by incorporating stochastic microbial dynamics and errors in measurements throughout the model. For Aim 2, we will develop Bayesian machine learning tools to predict host status from microbiome dynamics. The tools will learn easily interpretable, human-readable rules that predict host status from microbiome time-series data, and will be further extended to handle a variety of longitudinal study designs. For Aim 3, we will engineer our microbiome dynamics analysis software tools for optimal performance, ease-of- use, maintainability, extensibility, and dissemination to the community. In total, the proposed work will yield a suite of contemporary software tools for analyzing microbiome dynamics, with expected broad use and major impact. The software will allow investigators to answer important scientific and translational questions about the microbiome, including discovering which microbial taxa or their metagenomes are affected over time by perturbations such as changes in diet or invasion by pathogens; predicting the effects of these perturbations over time, including changes in composition or stability of the gut microbiota; and finding temporal signatures in multi-‘omic microbiome data that predict disease risk in the human host.