Distance-based Panomic Analytics for Microbiome Data

项目总结 我们研究微生物群的能力是由相同的技术实现的，这些技术使我们能够量化宿主 更深入、更精确的生理状态，包括用于基因组学的先进高通量测序 和转录组学；代谢组学、蛋白质组学和脂质组学的质谱学；以及流式细胞术 循环细胞群体的特征。宿主和微生物全基因组数据的集成是 未来生物医学发现的路线图。这类研究的一个例子是整合的人类微生物组 项目(IHMP)，该项目目前正在生成关于微生物及其宿主环境的全面数据，分三个阶段进行 不同的疾病(糖尿病、肠易激疾病、早产)。缺乏对这些问题的适当分析 数据及其验证和采用的陡峭曲线是社区主要关注的问题。面临的主要挑战 在全尺度微生物组数据集中的推论正在克服“维度灾难”。局部因果关系 事实证明，学习对于发现高维数据很有用，而基于远程的学习则是 多变量数据分析的一个很有前途的范例。我们建议将这些结合起来，以开发下一个 生成全套数据分析，并将这些工具直接提供给生物医学研究人员。这个 该项目的目标是：(1)开发基于距离的综合全息集成的分析；(2)开发 自上而下基于远程的子系统相互依赖学习方法。首要目标是 利用AIMS 1和AIMS 2中的方法开发面向用户的应用程序，并将其应用于几个现有的 研究如何生成全息数据。分析、应用程序和教育资源(案例研究和 由该项目产生的教程)将使生物医学界能够在 连贯而全面的方式。该项目产生的方法和工具将支持新的生物医学 发现。 好了！ 好了！

PROJECT SUMMARY Our ability to study the microbiomes is enabled by the same technologies that allow us to quantify the host physiological state at greater depth and precision, including advanced high-throughput sequencing for genomics and transcriptomics; mass spectrometry for metabolomics, proteomics, and lipidomics; and flow-cytometry for characterization of circulating cell populations. The integration of host and microbiome panomic data is the roadmap for future biomedical discoveries. One example of such studies is the Integrative Human Microbiome Project (iHMP), which is currently generating panomic data on microbes and their host environment in three different diseases (diabetes, irritable bowel disease, pre-term delivery). Lack of appropriate analytics for these data and a steep curve for their validation and adoption is a major concern for the community. The main challenge of inference in panomic-scale microbiome datasets is overcoming the ‘curses of dimensionality’. Local causal learning has proven useful for making discoveries with high-dimensional data, while distance-based learning is a promising paradigm for multivariate data analysis. We are proposing to combine these to develop the next generation of panomic data analytics and make these tools available directly to the biomedical investigators. The aims of this project are: (1) Develop analytics for distance-based omnibus panomic integration; (2) Develop methodology for top-down distance-based sub-system interdependence learning. The overarching goal is to develop user-facing applications utilizing the methodologies in Aims 1 and 2 and apply those in several existing studies generating panomic data. The analytics, applications, and educational resources (case studies and tutorials) resulting from this project will enable the biomedical community to study panomic-scale datasets in a coherent and comprehensive way. The methods and tools resulting from this project will support new biomedical discoveries. ! !