Construction and Retrieval of Highly Integrated Biological Databases

我们已经建立了一个分子相互作用的数据库，并开发了从中提取新的生物学知识的方法。对于这样一个数据库来说，能够综合处理化学信息以及基因组和蛋白质组学信息是很重要的。考虑到这一观点，我们取得了以下三个主要成果。BRITE数据库我们开发了BRITE数据库，以二元关系的形式存储直接和间接分子相互作用数据。主要包括酵母双杂交系统的蛋白互作数据、KEGG代谢途径中邻近酶的关系、KEGG调控途径中转录因子与其靶基因的关系。BRITE具有检索这些二进制数据并将其显示为网络的功能。从基因组数据库中提取本体我们实现了一个通用的数据间关系提取框架。使用关联规则发现方法（一种著名的数据挖掘方法），它可以快速发现给定条目集的公共和特定特征。接下来，我们通过构建一个来自基因组数据库的巨大词典来定义关键词与条目之间的关系。我们还构建了一个用于开发大型数据库的GRID环境。化学信息的数据库集成和网络拓扑分析我们开发了一种化合物二级结构的反应性表示格式和基于该格式的化学结构比较方法。我们还开发了一种推断酶反应中化学结构转化规律的算法，并应用该算法构建了一个反应物对数据库。该数据库进一步应用于预测新的酶反应途径。在拓扑分析方面，我们的目标是由化合物对和酶的关系形成的两个网络。我们对这两个网络和代谢网络中功能模块之间的关系有了新的认识。少

We have constructed a database of molecular interactions and developed methods for extracting novel biological knowledge from it. It is important for such a database to be able to handle chemical information as well as genomic and proteomic information as an integrated manner. Considering this viewpoint, we have achieved the following three main results.1.BRITE databaseWe have developed the BRITE database storing direct and indirect molecular interaction data as binary relations. It mainly consists of protein interaction data from yeast two-hybrid systems, neighboring enzyme relations in the KEGG metabolic pathway, and relationship between transcription factors and their target genes in the KEGG regulatory pathway. BRITE has a facility to retrieve these binary data and display them as a network.2.Ontology extraction from genome databasesWe implemented a general framework for extracting relationships among data. Using the association rule discovery method that is one of the well-known d … More ata mining methods, it quickly discovers common and specific features to a given set of entries. Next we defined relationship among keywords and entries by constructing a huge dictionary derived from genome databases. We also constructed a GRID environment for developing huge databases.3.Integration of chemical information into the database and analysis of network topologiesWe developed a representation format for secondary structures of chemical compounds in terms of reactivity and a method for comparing chemical structures based on the format. We also developed an algorithm to infer rules of chemical structure conversion in the enzyme reactions, and construct a database of reactant pairs by applying it. This database was further applied to a prediction of novel enzyme reaction pathways. Regarding the topology analysis, two networks created by the pairs of chemical compounds and enzyme relations were our targets. We obtained new insights into the relationship between the two networks and functional modules in the metabolic network. Less