Alignment Tools for the Conserved Domain Database

该项目涵盖了两个相互关联的软件开发工作。两者都旨在支持保存的域数据库CDD的管理员和用户，如项目LM000161-04中单独描述的那样。第一个努力需要开发和测试算法，旨在为不同的蛋白质序列产生准确的比对。第二项工作涉及开发交互式用户友好的工具，用于比对大量同源但不同的蛋白质序列片段，并用于识别这种蛋白质结构域超家族中的亚家族结构。 基于结构的比对算法的开发建立在我们团队早期在蛋白质线索方面的工作基础上，蛋白质线索是一套基于检测远距离同源关系的结构预测方法。这些方法通过结构模板“穿线”蛋白质序列，通过能量计算、使用接触电势和从模板的蛋白质家族获得的序列图谱来对备选比对进行评分。这些方法的成功在1998年的CASP3研讨会上得到了证明，在参加比赛的90多个国际小组中，NCBI团队被折叠识别授予结构预测方面的“第一名”。为了使这些方法适应CDD策展人所需的高吞吐量对齐，我们开发了用于线程的块对齐算法的更高效版本。早期的工作表明，这种方法产生的比对足够准确，可以识别保守的功能位点，并且相对于原始的穿线方法，信息损失最小。一种自动的多重比对精化算法，一次在一行迭代地应用基于结构的比对，已经经过了彻底的测试。它的性能表明它将有利于CDD精选工作，该算法正在CDD精选软件的主要组件Cn3D中实现。Cn3D已经包含块对齐工具的基本版本。该软件是CDD策展团队的日常使用软件，具有此扩展对齐功能的版本已被广泛分发。 进一步工作的重点是开发CDTree比对层次编辑系统，CDD馆长小组也在日常使用该系统。该软件在交互式软件包中实现了一套用于蛋白质家族分子进化分析的工具。它支持使用文献中的几种算法生成系统发育序列树，这些算法链接到生物分类树的显示和整个蛋白质结构域结构的总结。该软件还支持集成的“更新”程序，自动搜索每日更新的序列和结构数据库，以寻找每个家族的新成员，选择新的相似性和/或分类组的非冗余代表。CDTree子家族层次结构编辑器可与Cn3D比对编辑器无缝通信，使馆长用户能够轻松检测和纠正由比对或序列错误引起的“异常值”。虽然CDTree是为CDD策展人的需要而设计的，但它也可以用作简单的查看器，以直观的图形显示CDD家族层次结构中的序列、分类和功能多样性。 今年，CDTree已经过了彻底的测试，即将公开发布，以及一个新的、兼容的Cn3D版本。中国疾病控制与预防中心S网站服务已经准备好支持系统发育序列树的可视化，这为疾病控制与预防中心馆长定义的领域层次结构提供了证据。CDD S网络服务也准备好支持CD层次结构的下载和分析，使用CDTree作为用户S计算机上的辅助应用程序。CDTree？S集成？更新？PROCEDURE为NCBI？S PSI-BLAST程序提供了一个新颖的接口，该程序允许用户使用定制的位置特定分数矩阵搜索蛋白质数据库，之后提供了使用各种工具分析和提炼中间家族比对模型的机会。我们正在与BLAST小组合作，提供从BLAST结果页面直接启动CDTree的功能，这是由BLAST Web服务生成的。

This project covers two inter-related software development efforts. Both are intended to support curators as well as users of the Conserved Domain Database, CDD, as described separately in project LM000161-04. The first effort entails the development and testing of algorithms intended to produce accurate alignments for diverse protein sequences. The second effort involves development of interactive user-friendly tools for aligning large numbers of homologous but diverse protein sequence fragments, and for identifying the subfamily structure within such a protein domain super-family. Development of structure-based alignment algorithms builds upon our group's earlier work on protein threading, a set of structure prediction methods based on the detection of distant homologous relationships. These methods "thread" a protein sequence through a structural template, scoring alternative alignments by energy calculations, using contact potentials, and a sequence profile derived from the protein family of the template. The success of these methods was demonstrated at the 1998 CASP3 workshop, where the NCBI team was awarded "first place" in structure prediction by fold recognition, among over 90 international groups entering the competition. To adapt these methods to the high throughput alignment as needed by CDD curators we have developed more efficient versions of the block-alignment algorithm used in threading. Earlier work has shown that this method produces alignments accurate enough for identification of conserved functional sites, and that information loss relative to the original threading method is minimal. An automated multiple-alignment refinement algorithm, which iteratively applies structure-based alignment on one row at a time, has been thoroughly tested. Its performance suggests that it will benefit the CDD curation effort, and the algorithm is being implemented in Cn3D, a major component of CDD curation software. Cn3D already contains the basic version of the block alignment tool. This software is in daily use by the CDD curator team, and a version with this extended alignment functionality has been widely distributed. Further work has focused on development of the CDTree alignment hierarchy editing system, which is also in daily use by the CDD curator team. This software implements a suite of tools for molecular evolutionary analysis of protein families in an interactive package. It supports generation of phylogenetic sequence trees using several algorithms from the literature, linked to displays of organism taxonomy trees and summaries of overall protein domain architecture. The software also supports an integrated "update" procedure that automatically searches the daily-updated sequence and structure databases for new members of each family, selecting non-redundant representatives of new similarity and/or taxonomy groups. The CDTree subfamily hierarchy editor communicates seamlessly with the Cn3D alignment editor, allowing curator-users to easily detect and correct "outliers" caused by alignment or sequence errors. While designed for needs of CDD curators, CDTree may also be used as a simple viewer, showing in intuitive graphical displays the sequence, taxonomic and functional diversity within a CDD family hierarchy. This year CDTree has been exhaustively tested and public release is imminent, together with a new, compatible version of Cn3D. CDD?s web-services are ready to support visualization of phylogenetic sequence trees, which provide evidence for domain hierarchies as defined by CDD curators. CDD?s web-services are also ready to support download and analysis of CD hierarchies, using CDTree as a helper application on the user?s computer. CDTree?s integrated ?update? procedure provides a novel interface to NCBI?s PSI-BLAST program, which lets users search the protein database with customized position-specific score matrices, after providing an opportunity to analyze and refine the intermediate family alignment models with a variety of tools. We are working with the BLAST group to provide direct CDTree launch capability from BLAST results pages, as generated by the BLAST web services.