Conservation of Interaction Patterns in Protein Families

9506278 Godzik在实验确定蛋白质结构之前，通过计算机对蛋白质结构进行建模是一种非常宝贵的工具，有助于药物设计，突变研究和结构确定本身。 这项研究的主要目标是通过包括从已知蛋白质结构的新类型的分析中获得的额外信息来改进现有的蛋白质结构建模工具。 这整个方法可以最好地描述为在计算机上对蛋白质结构进行逆向工程。 蛋白质结构将被“分解”成相互作用的片段，并且通过利用“部分”库，它们将被重建。 为了在方法开发过程中获得经验并衡量进展，将已知结构建模为来自其结构家族的其他蛋白质的结构。 在“重建”已知结构中获得的经验将用于对一些真实的目标进行结构预测。 这项研究的主要科学信念是探索和利用一种新的、基于相互作用的蛋白质描述。 在这样的描述中，蛋白质被视为侧链之间相互作用的集合，而不是空间中的一组点;因此，负责折叠成特定结构的特征直接用于结构描述本身。 通过同源性建模获得的蛋白质模型的质量将通过使用通过这种新的视角获得的见解来提高，通过关注侧链相互作用，即使蛋白质骨架移动和滑动，侧链相互作用也会随着序列的变化而保持不变。 本研究的目标将分两个阶段实现。 在第一阶段，传统的建模技术将得到补充，从蛋白质家族的分析获得的额外信息。 在第二阶段，将开发一种新的建模方法，其中预测的蛋白质结构将由内而外地构建，围绕相互作用的蛋白质簇，使用来自其他蛋白质的结构片段。 为协助实现这些目标，将开发一套必要的工具。 一些具体的目标，包括建立与目标具有不同同源性水平的蛋白质模型，将用于判断研究进展。 ***

9506278 Godzik Modeling of protein structures by computer, before it is determined experimentally, is an invaluable tool, aiding drug design, mutation studies and often structure determination itself. The main goal of this research is to improve existing tools for the modeling of protein structures by including extra information obtained from the new type of analysis of regularities in known protein structures. This whole approach can be best described as the reverse- engineering of protein structures on the computer. Protein structures will be "decomposed" into interacting fragments, and, by utilizing a library of "parts", they will be rebuilt. To gain experience as well as to measure progress during the method development, known structures will be modeled onto structures of other proteins from their structural families. Experience gained in the "rebuilding" of known structures will be used to make structural predictions for a number of real targets. %%% The main scientific trust in this research is to explore and utilize a new, interaction based, description of proteins. In such a description, a protein is seen as a collection of interactions between side chains, rather than as a set of points in space; thus, the very features responsible for the folding to a particular structure are directly used in the structure description itself. The quality of protein models obtained by homology modeling will be improved by using insights obtained through this new perspective, by focusing on side-chain interactions which remain constant with sequence changes even as protein backbone shifts and slides. The goals of this research will be achieved in two stages. In the first stage, traditional modeling techniques will be supplemented by additional information obtained from the analysis of protein families. In the second stage, a new modeling approach will be developed, where the predicted protein structure will be built inside-out, around interacting clusters of res idues, using fragments of structures from other proteins. To assist in achieving these goals, a set of necessary tools will be developed. A number of specific objectives, including building models of proteins with various levels of homology to the target, will be used to judge the progress of the research. ***