Collaborative Research: Statistical Methodology for Network based Integrative Analysis of Omics Data

该项目的总体目标是基于基因、转录本、蛋白质和代谢物的协调活动来描绘可能作为治疗靶点的途径，以及创建基于Omics的生物标记物小组，用于疾病的早期检测和预后。具体地说，我们解决了具有网络结构的多源高维生物数据的集成和分析问题。事实证明，不同分子间的相关性相对较低，而从单一分子间获得的信息往往噪音很高，甚至不完整。因此，需要开发用于集成来自不同OMICS平台的多个数据集的高级模型和技术，同时明确考虑隔间内和隔间交互的可用信息。由于它们在复杂疾病的发生和发展中的作用，因此特别强调途径分析和丰富。将寻求以下方向：(1)开发基于网络的方法，整合来自多个OMICS平台的数据，用于路径分析和丰富。(2)大规模网络综合模型估计的快速算法研究。研究相关的推理问题，研究所提出的估值器的性质以及它们对所用网络信息的噪声水平的稳健性。(3)引入新的超图模型来评估通路的差异活动，这些通路利用关于底层网络的结构和准确性的不同程度的信息。(4)开发了一种基于扰动P值的新方案，用于检测有助于发现生物标记物的途径的活跃成员。(5)将拟议的方法转化为易于使用的软件工具。拟议的研究计划将产生三方面的影响：方法论、科学性和教育性。在方法学方面，基于该项目的研究将导致：(A)开发一个基于不同模型的综合框架，用于评估通路的不同活性，该模型集成了来自多个Omics平台的数据，并利用关于底层网络的结构和精度的不同程度的信息；(B)系统地了解大规模(广义)混合线性模型中涉及的计算问题；以及(C)基于扰动P值的新方案，用于识别成为治疗药物潜在靶点的通路的活跃成员。科学认识的加强将在应用层面上产生切实的影响。许多拟议的方法已经被用于高维基因组、蛋白质组和代谢组数据的分析，重点是识别不同疾病(主要是癌症)状态下的活动通路(子网络)。此外，一些新的实验正在设计阶段，这些实验将利用本项目中提出的一些先进模型和技术。这项提议的另一个关键方面是开发一个易于从业者使用的开源软件，构建在一个领域独立的工作流管理系统内。这允许用户以一种简单和透明的方式通过添加他们自己的功能和计算工具来增强软件。这项研究议程所产生的新的方法程序将通过跨学科的互动和协作以及在会议和专门讲习班上的发言向相关科学界传播。最后，在教育方面，该项目的材料将为在私人投资机构指导下工作的博士生提供研究课题；因此，它将在培养未来的量化科学家方面发挥重要作用。

The overarching goal of this project is to delineate pathways-based on coordinated activity of genes, transcripts, proteins and metabolites, that could potentially serve as therapeutic targets, as well as create Omics based biomarker panels for early detection and prognosis of disease. Specifically, we address the problem of integration and analysis of multiple sources of high dimensional biological data with network structure. It is a well documented fact that correlation between different molecular compartments is relatively low, while the information derived from a single compartment is often highly noisy or even incomplete. Hence, there is a need to develop advanced models and techniques for integrating multiple data sets from diverse Omics platforms, while taking explicitly into consideration the available information of interactions within and between compartments. Particular emphasis is placed on pathway analysis and enrichment due to their role in complex diseases onset and progression. The following directions will be pursued: (1) Development of network based methods that integrate data from multiple Omics platforms for pathway analysis and enrichment. (2) Development of fast computational algorithms for estimating large scale network based integrative models. Investigation of associated inference problems and study of properties of proposed estimators together with their robustness to the noise levels of the network information employed. (3) Introduction of novel hypergraph models for assessing differential activity of pathways that utilize different degrees of information about the structure and accuracy of the underlying network. (4) Development of a novel scheme based on perturbed P-values for detecting active members of pathways that would aid in biomarker discovery. (5) Implementation of the propose methodology into an easy to use software tool.The proposed research program will have a three-pronged impact: methodological, scientific and educational. On the methodology front, the research based on this project will lead (a) to a developing a comprehensive framework for assessing differential activity of pathways based on different models that integrate data from multiple Omics platforms and utilize different degrees of information about the structure and accuracy of the underlying network, (b) a systematicunderstanding of the computational issues involved in large scale (generalized) mixed linear models;and (c) a novel scheme based on perturbed P-values for identifying active members of pathways that become potential targets for therapeutic drugs. The enhanced scientific understanding will provide tangible impact at the level of applications. A number of the proposed methods have already been used in the analysis of high dimensional genomic, proteomic and metabolomic data with emphasis on identifying active pathways (subnetworks) in different disease (primarily cancer) states. Further, a number of new experiments are in the design stage that would utilize some of the advanced models and techniques proposed in this project. Another key aspect of this proposal is the development of an easy to use by practitioners open source software, built within a domain independent workflow management system. This allows users to enhance the software by adding their own functionality and computational tools in an easy and transparent manner. The novel methodological procedures ensuing from this research agenda will be disseminated to the relevant scientific communities, both via inter-disciplinary interaction and collaboration and through presentations at conferences and specialized workshops. Finally, on the educational front, the material from the project will provide research topics for doctoral students working under the supervision of the PIs; it will therefore play an important role in the training of future quantitative scientists.