A Software and Data Resource for Multidomain Homology and Protein Classification

卡耐基梅隆大学获得了一项资助，用于开发多域序列家族同源性检测和分类的通用软件和数据资源。可靠的同源性鉴定对于基因注释、功能预测和比较基因组学的准确性至关重要。同源性鉴定是单结构域蛋白的一个很好的研究问题，但对于复杂的多结构域蛋白仍未解决。该项目将开发手工验证数据集资源；高精度的多域同源物和正交物检测方法多域族自动分类方法研究还有一个公开可用的数据库，带有基于网络的界面和可视化工具，用于对数据进行探索性分析。该平台的一个关键组成部分是利用序列相似网络的局部结构来推断多域家族关系的新颖而准确的方法。这将为研究区域洗牌的进化过程提供一个平台；多结构域家族特异性进化过程中基因复制、结构域复制、插入、缺失和重排的历史以及这些家族在当代生物体中所扮演的功能角色。准确识别多结构域同源物的方法将促进基因功能预测和比较基因组学的广泛应用，从而为必要的研究基础设施做出贡献。从科学上讲，本研究解决了分类的基本问题，这些问题是广泛应用的基因组分析的基础，包括蛋白质功能预测，特别是复杂的多结构域蛋白质。这类蛋白质在脊椎动物基因组中特别有趣，因为它们与发育、神经功能、组织修复和免疫系统有关。该提案的研究主题将被纳入匹兹堡超级计算中心关于发展生物信息学计划的研讨会的讲座中，这是一个为期两周的课程，旨在为MARC（少数民族获得研究职业）计划的教师在当地校园教授生物信息学课程做准备。本科生将通过“进化与生命史”课程项目参与个别多域家族的研究。

Carnegie Mellon University is awarded a grant to develop general purpose software and data resource for homology detection and classification of multi-domain sequence families. Reliable homology identification is essential for accuracy in gene annotation, function prediction and comparative genomics. Homology identification is a well studied problem for single domain proteins, but remains unsolved for complex, multi-domain proteins. The project will develop a resource of handcurated validation data sets; high accuracy methods for detection of multi-domain homologs and orthologs; methods for automated classification of multi-domain families; and a publicly available database with a web-based interface and visualization tools for exploratory analysis of the data. A key component of the proposed platform is a novel and accurate method for inferring multi-domain family relationships by exploiting the local structure of the sequence similarity network. It will produce a platform for studying the evolutionary processes of domain shuffling; the histories of gene duplications and domain duplications, insertions, deletions and rearrangements in the evolution of specific of multi-domain families; and the functional roles these families play in contemporary organisms. Accurate methods to identify multi-domain homologs will enhance a broad range of applications in gene function prediction and comparative genomics, thus contributing to essential research infrastructure. Scientifically, the proposed research addresses basic problems of classification that are fundamental to a broad range of widely used genomic analyses, including protein function prediction, especially for complex, multi-domain proteins. Such proteins are of particular interest in vertebrate genomes, where they are implicated in development, neural function, tissue repair and the immune system. The research themes of the proposal will be incorporated into lectures for the Pittsburgh Supercomputing Center's workshop on Developing Bioinformatics Programs, a two-week course aimed at preparing MARC (Minority Access to Research Careers) program faculty to teach bioinformatics courses at their local campuses. Undergraduates will participate in research on individual multi-domain families through course projects in ''Evolution and the History of Life''.