Building motif lexicons

描述（由申请人提供）：随着大量基因组的完整测序和蛋白质组的注释，生物学的下一个主要挑战之一是了解编码蛋白质的功能和整合（NIH路线图重点领域）。破译蛋白质功能是一个非常耗时，昂贵的过程，这反映在具有良好功能的蛋白质的比例不成比例地低。将已建立的功能外推到新蛋白质的一种方法是预测短基序。短基序靶向蛋白质进行翻译后修饰，运输到细胞区室，并与其他蛋白质或分子结合。我们的跨学科团队已经建立了Minimotif Miner（MnM），这是一个短基序数据库和平台独立的网络工具，用于识别蛋白质查询中的基序共有序列，从而识别潜在的新蛋白质功能（http：//mnm.engr.uconn.edu/）。MnM还可用于开发特定突变如何导致人类疾病的新假设，并确定开发治疗药物，抗生素，杀虫剂和抗病毒剂的推定靶标。尽管MnM和其他基序资源的效用，预测功能基序仍然有两个主要的限制，我们在这个建议中解决。1)为了减少对基序的假阳性预测，我们创建了一种新的语言，使我们能够考虑基序的三维结构保守性。对于每个基序，我们将通过结合实验数据和蛋白质数据库中的基序结构数据来构建特定的基序定义。我们还将通过分子动力学模拟来确定可以形成所观察到的基序结构的序列排列。2)为了建立一个更全面的主题数据库，我们将使用人工智能挖掘PubMed。专家系统将使用自动化文献筛选、文档摘要和基序识别效率评分来从PubMed中提取大多数已知基序。解决这些限制将大大增加短基序预测的实用性。

DESCRIPTION (provided by applicant): With the complete sequencing of numerous genomes and the annotation of proteomes, one of the next major challenges in biology is to understand the functions and integration of the encoded proteins (a NIH Roadmap area of emphasis). Deciphering protein function is a very time consuming, expensive process, as reflected by the disproportionately low percentage of proteins with well-established functions. One approach for extrapolating established functions to new proteins is to predict short motifs. Short motifs target proteins for post-translational modification, trafficking to cellular compartments, and binding to other proteins or molecules. Our cross-disciplinary team has built Minimotif Miner (MnM), a short motif database and platform- independent web-tool that identifies motif consensus sequences in protein queries and thus potential new protein functions (http://mnm.engr.uconn.edu/). MnM can also be used to develop new hypotheses of how specific mutations cause human disease and to identify putative targets for the development of therapeutic drugs, antibiotics, insecticides, and antiviral agents. Despite the utility of MnM and other motif resources, prediction of functional motifs still has two major limitations, which we address in this proposal. 1) To reduce the false-positive prediction of motifs, we have created a new language that allows us to consider the 3- dimensional structural conservation of motifs. For each motif, we will build specific motif definitions by combining experimental data with data from motif structures in the Protein Data Bank. We will also determine the sequence permutations that can form the observed motif structure by using molecular dynamic simulations. 2) To build a more comprehensive motif database we will use artificial intelligence to mine PubMed. The expert system will use automated literature screening, document summarization, and motif identification efficiency score to extract the majority of known motifs from PubMed. Addressing these limitations will vastly increase the utility of short motif prediction.