Template-Based Protein Structure Prediction Beyond Sequence Homology

“这项奖励是根据2009年美国复苏和再投资法案（公法111-5）资助的。”普渡大学获得了一项资助，用于开发用于构建蛋白质结构模型的新型计算算法和软件，该模型可以识别和利用非常遥远的模板结构进行建模。由于结构基因组学项目产生的蛋白质结构越来越多，因此充分利用这些结构作为模板进行建模对蛋白质结构预测方法具有重要意义。然而，现有方法在搜索模板结构和计算模板比对时过于依赖序列相似性。提出的项目旨在通过在计算目标模板对齐中广泛使用结构信息和引入使用先进从头算折叠的改进程序来推动模板搜索和结构建模的限制。该项目的主要创新包括(1)新颖的线程算法，对模板蛋白中的残馀接触进行概率处理，(2)使用次优对齐和多个模板来提高模型的准确性，以及(3)使用从头算折叠方法来改进模型，该方法考虑了模型每个区域的估计误差。对目标蛋白建模模板识别的改进将利用实验解决结构的价值，并通过减少建模生物体蛋白质组结构所需的模板结构的数量，显著降低结构基因组学项目的成本。该项目是与蛋白质结构预测领域的国际专家合作进行的。生物科学和计算机科学的研究生和本科生将在几个系的交叉课程中进行培训。普渡大学现有的几个招收少数民族学生和本科生的项目将有助于广泛参与该项目。总的来说，拟议的项目利用了普渡大学？在跨学科计算生命科学与工程方面的努力。在这笔资金下开发的软件工具可以从PI的实验室网站http://dragon.bio.purdue.edu/访问。

"This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5)."Purdue University is awarded a grant to develop novel computational algorithms and software for building protein structure models, which can identify and utilize very distantly related template structures for the modeling. Considering the growing number of protein structures produced by structural genomics projects, it is important for protein structure prediction methods to take full advantage of those structures as templates in modeling. However, existing methods rely too heavily on sequence similarity in searching template structures and computing alignments with templates. The proposed project is aimed to push the limit of template searching and structure modeling by extensive use of structural information in computing target-template alignments and by introducing a refinement procedure, which uses an advanced ab initio folding. Key innovations of this project include (1) novel threading algorithms with a probabilistic handling of residue contacts in template proteins, (2) use of suboptimal alignments and multiple templates to improve accuracy of a model, and (3) use of an ab initio folding method for refining the model, which takes estimated errors at each region of the model into account. Improvement of recognition of a template for modeling a target protein will leverage the value of experimentally solved structures and will also lead to a significant reduction of the cost of structural genomics projects by reducing the number of template structures needed for modeling the structures of proteome of organisms.The proposed project is carried out in international collaboration with world experts in the protein structure prediction field. Graduate and undergraduate students in biological sciences and computer science will be trained in cross-listed courses among several departments. Several existing programs at Purdue for recruiting minority students and undergraduate students will contribute to broad participation in the project. Overall the proposed project leverages Purdue University?s efforts in interdisciplinary computational life science and engineering. Software tools developed under this funding may be accessed from the PI's lab website at http://dragon.bio.purdue.edu/.