EAGER: Graph-Based Theoretical Models and Mining Algorithms for Bioinformatics Data Analysis

EAGER：生物信息学数据分析项目的基于图的理论模型和挖掘算法摘要图出现在令人惊讶的多种学科中，从计算机网络到社会学、生物学、生态学等等。近年来，在生物信息学应用中产生了大量的数据，并最终以图形格式表示。有些存储为单个大图形，如蛋白质-蛋白质相互作用网络，而另一些存储为图形数据库中的一组图形对象，如药物设计中的化合物和生物信息学中的蛋白质2D/3D结构。本提案的主要目标是开发新的基于图的理论模型和算法来分析以单个大图或图数据库表示的生物信息学数据集，重点是(1)开发基于图分解和压缩的图数据库中有效的结构模式发现方法，以及(2)开发用于大规模无尺度网络图的挖掘算法。重点研究复杂网络中层次结构的混合模型、统计推理和缺失环节预测。这些算法将通过分析生物信息学研究中产生的数据集来验证。该研究将促进对生物信息学和数据挖掘算法在图数据集中的研究和开发的理解。它还将扩大应用范围，推动图论、生物信息学和数据挖掘的新领域。这项研究工作被认为是？高风险高回报？因为它需要开发新的理论图模型和算法，与主流的随机图理论模型和算法完全不同。智力优势：本提案中开发的技术和方法建立在生物信息学，图理论，数据挖掘和数据库管理的最先进方法的基础上。这项研究将使人们更好地理解在科学数据集中设计有效的基于图形的模型、算法和方法所涉及的问题。拟议的项目将设计和开发广泛的新型数据分析算法和方法，包括结构模式匹配和发现，以及挖掘大规模无尺度网络图。更广泛的影响：该提案将解决诸如生物信息学、药物设计、分子生物学等科学应用领域的数据分析中的广泛问题。从这项研究中发展出来的基于图的模型和算法都是计算机科学的核心，并且是可推广的，并将在其他领域公开使用，包括社会学和流行病学中的个人或社会联系，信息科学中的作者共同引用，计算机科学和信息技术中的互联网和万维网。本课题的研究成果可以为生物网络的建模和仿真机制提供更高效、更有效的建模和仿真机制，并为其他科学和工程领域的应用提供先进的数学能力。PI和Co-PI坚定致力于整合研究和教育，促进多样性，牢固的工业伙伴关系和广泛传播研究成果。拟议的研究领域有助于在许多层面上提高学生的科学好奇心。学生将接触到最新的研究成果。该提案的研究和教育计划都是高度跨学科的，吸引了来自不同研究领域的学生和教师，并借鉴了许多研究领域的工作。特别是，这些计划有助于强调综合生物信息学、图论和数据挖掘在为生物信息学研究创建下一代基于图的模型、算法和工具方面的好处。有了这些模型、算法和工具，生物信息学研究人员将发现更多有意义和相关的知识/模式，并使他们能够更好地理解和解释他们在调查中收集的数据集。学生将获得理解、分析和挖掘越来越多的生物信息学数据集的能力。

EAGER: Graph-based Theoretical Models and Mining Algorithms for Bioinformatics Data Analysis Project SummaryGraphs show up in a surprisingly diverse set of disciplines, ranging from computer networks to sociology, biology, ecology and many more. In recent years, huge amounts of data are generated and ultimately represented in graph format in the bioinformatics applications. Some are stored as a single large graph such as a protein-protein interaction network, while others are stored as a set of graph objects in a graph database such as chemical compounds in drug design and protein 2D/3D structures in bioinformatics. The main goal of this proposal is to develop novel graph-based theoretical models and algorithms to analyze bioinformatics data sets represented as either a single large graph or a graph database, focusing on the (1) development of efficient structure pattern discovery methods in graph databases based on graph decomposition and compression, and (2) development of mining algorithms for large scale-free network graphs, focusing on developing mixture model and statistical inference of hierarchical structure and missing link prediction in complex network. These algorithms will be validated by analyzing data sets arising in bioinformatics research. The proposed research will advance the understanding of bioinformatics and data mining algorithm research and development in graph data sets. It will also expand the application scope and push new frontiers for graph theory, bioinformatics and data mining. This research work is considered ?high risk high payoff? because it needs to develop novel theoretical graph-model and algorithms, radically different from the dominant random graph theory models and algorithmsIntellectual merit: The techniques and methods to be developed in this proposal build on state-of-the-art methods in bioinformatics, graph theory, data mining and database management. The research will result in improved understanding of the issues involved in designing efficient graph-based models, algorithms and methods in scientific data sets. The proposed project will design and develop a wide-range of novel data analysis algorithms and methods including structure pattern matching and discovery, and mining large scale-free network graphs. Broader impacts: The proposal will address a broad range of problems in the data analysis of scientific application domains such as bioinformatics, drug design, molecular biology, etc. Both the graph-based models and algorithms developed from this research are central to the computer science, and are generalizable and will be made publicly available for use in other domains, including personal or social contacts in sociology and epidemiology, author-co-citations in information science, the Internet and the World Wide Web in computer science and information technology. Research outcomes from this proposal can lead to more efficient and effective modeling and simulation mechanism of the biological network and advanced mathematical capabilities that are applicable through other science and engineering domain. The PI and Co-PI have a strong commitment to the integration of research and education, promotion of diversity, strong industrial partnership and broad dissemination of research results. The proposed research area lends itself to raising the scientific curiosity of students at many levels. Students will obtain significant exposure to the latest research. Both the research and education plans of the proposal are highly interdisciplinary, engaging students and faculty from various research areas and drawing from work on numerous fields of study. In particular, the plans serve to highlight the benefits of synthesizing bioinformatics, graph theory and data mining in creating the next generation of graph-based models, algorithms and tools for bioinformatics research. Armed with these models, algorithms and tools, bioinformatics researchers will discover more meaningful and pertinent knowledge/patterns and enable them to have a better understanding and interpretation of the data sets collected in their investigation. Students will gain an appreciation for the ability to understand, analyze and mine the growing range of bioinformatics data sets.