Computational Studies of Virus-host Interactions Using Metagenomics Data and Applications

用元基因组学数据和计算机研究病毒与宿主的相互作用 应用 摘要：病毒在几乎所有的生态环境中无处不在，包括人类 人体、水、土壤等，对人体微生物群的正常功能起着重要作用。 许多病毒已被证明与人类疾病有关。然而，我们的 对病毒在生态群落中的作用的了解非常有限。近期 技术和计算的进步使人们有可能深刻地理解 病毒在公共卫生和环境中的作用。来自不同国家的元基因组学研究 环境，包括人类微生物组计划(HMP)、全球海洋和地球 微生物组项目产生了大量的短读取数据。病毒存在于 这些元基因组数据集和它们的宿主大多是未知的。在这项提案中， 研究人员将开发识别病毒序列的计算方法 从元基因组数据集和研究病毒与宿主的相互作用。用于识别 来自元基因组样本的病毒序列，使用单词模式的新统计测量 将首先被开发出来。二是统一的朴素贝叶斯综合方法 将研究来自词型、基因方向性和基因注释的信息。第三, 从后基因组中鉴定出的病毒序列将被进一步组装以构建 使用将由研究人员开发的一种新的装箱方法来完成病毒基因组。 最后，剩余的读数将被分配到相应的存储箱。对于病毒的研究- 宿主相互作用，估计病毒-宿主相互作用可靠性的计算方法 来自高通量的实验将首先开发出来。然后是机器学习方法 将被开发来预测感染某些宿主的病毒。最后，建立了网络逻辑回归模型 将开发一种方法来预测病毒与宿主的相互作用。这些计算方法 用于识别病毒序列和预测病毒与宿主的相互作用 以一个公共的肝硬变和一个独特的元基因组数据集来了解元基因组是如何 随健康状态变化，识别与疾病相关的病毒和病毒-宿主交互作用 利用细菌、病毒和病毒与宿主的相互作用，准确预测疾病状态。 开发的计算方法还将用于分析来自 基于不同地点的塔拉海洋数据和独特的时间序列数据来理解 环境因素如何影响病毒丰度和病毒与宿主的相互作用。其中一些 预测将得到实验验证。从提案派生的软件将是 编写并免费分发给科学界。

Computational Studies of Virus-host Interactions Using Metagenomics Data and Applications Summary: Viruses are ubiquitous in almost every ecological environment including the human body, water, soil, etc. They play important roles in the normal function of human microbiome. Many viruses have been shown to be associated with human diseases. However, our understanding of the roles of viruses in ecological communities is very limited. Recent technological and computational advances make it possible to have a deep understanding of the roles of viruses in public health and the environment. Metagenomics studies from various environments including the human microbiome projects (HMP), global ocean, and the earth microbiome projects have generated large amounts of short read data. Viruses are present in most of these metagenomic data sets and their hosts are unknown. In this proposal, the investigators will develop computational approaches for the identification of viral sequences from metagenomic data sets and for the study of virus-host interactions. For the identification of viral sequences from metagenomics samples, novel statistical measures using word patterns will first be developed. Second, a unified naïve Bayesian integrative approach by combining information from word patterns, gene directionality, and gene annotation will be studied. Third, the identified viral sequences from metagenomes will be further assembled to construct complete viral genomes using a novel binning approach to be developed by the investigators. Finally, the remaining reads will be assigned to the corresponding bins. For the study of virus- host interactions, computational methods to estimate the reliability of virus-host interactions from high-throughput experiments will first be developed. Then machine learning approaches will be developed to predict viruses infecting certain hosts. Finally, a network logistic regression approach will be developed to predict virus-host interactions. These computational approaches for the identification of viral sequences and for predicting virus-host interactions will be applied to a public liver cirrhosis and a unique metagenomics data set to understand how metagenomes change with health status, identify viruses and virus-host interactions associated with disease status and accurately predict disease status using bacteria, viruses and virus-host interactions. The developed computational methods will also be used to analyze metageomic data from various locations based on the TARA ocean data and a unique time series data to understand how environmental factors affect virus abundance and virus-host interactions. Some of the predictions will be experimentally validated. Software derived from the proposal will be developed and freely distributed to the scientific community.