Similarity-Based Indexing and Integration of Protein Sequence and Structure Databases

俄亥俄州立大学获得拨款，用于开发数据库索引和相似性搜索技术，以管理、分析和集成蛋白质序列和结构数据库。在基因组和蛋白质组数据库中搜索相似的序列和结构是生物信息学的一项基本任务。随着可用数据规模的迅速增加，建立索引方案以有效实现序列数据和结构数据的集成维护和查询至关重要。为了应对这一挑战，该项目对两种类型的数据使用统一的主题：提取关键特征并将其映射到紧凑的特征向量空间中，以方便构建具有敏感、准确和高效查询能力的集成索引结构。对于序列数据，该项目将开发涉及氨基酸的理化特性并检测低水平相似性的新颖特征提取。对于结构数据，该项目将开发使用接触图和空间图案捕获局部结构图案的方法。在这两种情况下，都将构建特征的紧凑表示，以及对它们进行索引的有效结构。该方法将分子的生化蛋白质纳入特征提取中，以发现蛋白质的功能位点并返回生物学相关的查询结果。最后，基于统一的特征表示和索引框架，该项目将开发在各个层面上有效集成序列和结构数据的方法。结合序列和结构数据的整体方法将有助于克服各自的局限性，并提供更准确的查询结果。该项目的成果将有利于自然科学和健康科学的广泛应用领域，包括：比较和功能基因组学、蛋白质建模和设计、药物开发以及预防和个性化医学。该项目开发的软件将促进大规模全基因组研究项目，这些项目需要对可用序列和结构数据库进行迭代和交互式查询。开发的新颖表示和敏感主题提取方法也适用于生物数据可视化、分类和多重比对问题。该软件和该项目的结果将在网站上提供：http://bio.cse.ohio-state.edu。

The Ohio State University is awarded a grant to develop database indexing and similarity search technologies to manage, analyze, and integrate protein sequence and structure databases. Searching for similar sequences and structures in genomic and proteomic databases is a fundamental task in bioinformatics. As the size of the available data increases rapidly, it is essential to build indexing schemes so that integrated maintenance and querying of both sequence and structure data can be achieved effectively. To address this challenge, this project uses a unified theme for both types of data: extracting key features and mapping them into compact feature vectors spaces to facilitate construction of integrated index structures with sensitive, accurate, and efficient querying capabilities. For the sequence data, the project will develop novel feature extraction that involve physiochemical properties of the amino acids and detect low level of similarities. For the structural data, the project will develop methods to capture local structural motifs using contact maps and spatial motifs. In both cases, compact representation of features will be constructed, as well as efficient structure to index them. The approach incorporates biochemical proteins of molecules into feature extraction to discover functional sites of proteins and to return biologically relevant query results. Finally, based on the unified feature representation and indexing framework, the project will develop methods to integrate sequence and structure data effectively at various levels. A holistic approach combining sequence and structure data would help to overcome the limitations of each, and provide more accurate query results. The results of the project will benefit a wide range of application areas in natural and health sciences, including: comparative and functional genomics, protein modeling and design, drug development, and preventative and personalized medicine. Software developed in this project will facilitate large-scale genome-wide research projects which require iterative and interactive querying of available sequence and structure databases. The novel representations and sensitive motif extraction methods developed are also applicable to biological data visualization, classification, and multiple alignment problems. The software and the results of this project will be available at the website: http://bio.cse.ohio-state.edu.