An integrated approach to infer and validate domain-domain interactions in protei

描述（由申请人提供）：蛋白质通过其组成结构域相互通信。域-域相互作用的知识将成为蛋白质功能注释的一部分，这对生物医学研究界非常有用。尽管有大量的蛋白质相互作用数据，但我们目前对结构域-结构域相互作用的知识非常有限，因为大多数蛋白质-蛋白质相互作用数据都是“二元”数据（即，相互作用要么被发现，要么没有被发现），这并没有揭示哪两个结构域是相互作用的。用实验的方法来确定所有的域与域之间的相互作用既繁琐又不可行。为了利用大量的蛋白质相互作用数据，可以利用计算方法从蛋白质-蛋白质相互作用数据推断域-域相互作用。该项目的具体目标是：(i)开发一种新的计算方法，用于从蛋白质-蛋白质相互作用数据推断生物学相关的域-域相互作用，以及（ii）通过酵母双杂交筛选实验验证预测的相互作用。总之，综合评分算法将采用一种新的评分特征组合，从所有理论上可能的交互池中准确地推断出潜在的域-域交互。将利用来自多个来源和多个物种的实验导出的蛋白质-蛋白质相互作用数据集，以确保最大限度地覆盖域-域相互作用。该方法的性能将通过对iPfam数据库中实验已知域相互作用的预测进行测试来评估。此外，将使用酵母双杂交筛选对100种相互作用进行潜在的正相互作用和负相互作用的直接实验验证。35个得分最高和35个得分最低的预测将被实验验证，以评估这些预测的质量。此外，我们建议测试在一组代表肿瘤抑制因子的DNA修复蛋白核心中发现的大约30个新的结构域-结构域相互作用。在这个项目中产生的方法和数据可以帮助理解蛋白质-蛋白质相互作用的功能基础，在系统生物学研究中具有多种含义。

DESCRIPTION (provided by applicant): Proteins communicate with each other using their constituent domains. Knowledge of domain-domain interactions will become part of a protein's functional annotation that is widely useful to the biomedical research community. Despite the availability of a vast amount of protein interaction data, our current knowledge on domain-domain interactions is very limited, because most of the protein-protein interaction data exist as `binary' data (i.e., interaction is either found or not found) that does not reveal which two domains are interacting. Determining all domain-domain interactions using experimental means is tedious and unfeasible. To take advantage of the vast amount of protein interaction data, computational methods can be exploited for inferring domain-domain interactions from protein-protein interaction data. Specific aims of this project are: (i) To develop a novel computational method for inferring biologically relevant domain-domain interactions from protein-protein interaction data, and (ii) To experimentally validate predicted interactions using yeast two-hybrid screens. In summary, a novel combination of scoring features will be employed in an integrated scoring algorithm, to accurately infer potential domain-domain interactions from the pool of all the theoretically possible interactions. Experimentally derived protein-protein interaction datasets from multiple sources and from multiple species will be utilized to ensure maximum coverage of domain- domain interactions. The performance of this method will be evaluated by testing the predictions against experimentally-known domain interactions in the iPfam database. Additionally, direct experimental validation of potential positive and negative interactions will be carried out for 100 interactions using the yeast two-hybrid screens. A selected list of 35 highest-scoring and 35 lowest-scoring predictions will be experimentally validated to assess the quality of these predictions. Additionally, we propose to test about 30 novel domain- domain interactions found in a core set of DNA repair proteins representing tumor suppressors. The methodology and data generated in this project can help understand the functional basis of protein-protein interactions, with a multitude of implications in systems biology research.