Alignments and Improved Refinements for High-Accuracy Protein Structure Modeling

描述(由申请人提供)：这份题为《用于高精度结构建模的正确比对和改进的精细化》的提案紧跟NIH RFA定义的两个目标，即(1)获得近同系物的晶体结构质量模型(超过30%的序列一致性)和(2)建立远端同源物的高精度模型(低至10%的序列一致性)。我们认为，要实现这些目标，将需要(A)开发新的方法来整合蛋白质序列家族中所有可用的生物和结构信息，以改进目标蛋白质序列与已知结构模板的比对；(2)对蛋白质结构可访问的构象空间进行采样的新方法；以及(3)提供精确的近天然蛋白质结构模型的新方法。我们将通过以下具体目标实现这些目标。 (1)通过组合广泛的不同信息源来生成目标蛋白序列和结构模板之间的准确比对。这一目标的关键是将这些数据整合到一个统一的框架中。我们将发展残基位置注释(RPA)的概念来进行同源建模，其中序列中的不同位置根据来自多个序列比对的属性在建模过程中产生不同的影响。我们将使用平均场最小化的框架来设计一个新的比对方案，它包含了不同类型的约束，甚至是非相加的。 (2)建立高精度的目标蛋白质模型。基于目标序列和已知结构的模板蛋白序列之间的给定序列比对的同源性建模通常包括两个步骤：环路构建，确定比对中的插入和缺失区域，以及侧链建模。我们将根据我们为这两个问题制定解决方案的丰富经验，提出绕过这些困难的新办法。 (3)改进初始模型，通过结构细化生成晶体结构质量模型。我们将遵循三个方向(A)改进力场的最小化和分子动力学，包括量子力学项，(B)改进的隐式溶剂模型的最小化和分子动力学，和(C)合作多体能量项的能量最小化。 (4)将本提案中开发的所有计算机程序组织成一个以用户为中心的包，包括视觉计算工具，以使整个生物界都能访问它。

DESCRIPTION (provided by applicant): This proposal entitled "Correct Alignments and Improved Refinements for High-Accuracy Structure Modeling" follows closely the two goals defined by the NIH RFA, namely, (1) getting crystal structure quality models for close homologs (more than 30% sequence identity) and (2) building high accuracy models for remote homologs (as low as 10% sequence identity). We believe that reaching these goals will require the development of (a) new approaches to integrate all biological and structural information available on a protein sequence family to improve the alignment of the target protein sequence to a known structural template, (2) new methods for sampling the conformational space accessible to a protein structure, and (3) new methods that provide accurate refinements of near native protein structural models. We will achieve these goals through the following specific aims. (1) Generate accurate alignments between the target protein sequence and a structural template by combining a wide range of different sources of information. Essential to this aim is the integration of these data into a unified framework. We will develop the concept of residue position annotation (RPA) for homology modeling, in which different positions in the sequence have different impact in the modeling procedure depending on their properties, derived from multiple sequence alignments. We will use the framework of mean field minimization to design a new alignment package that incorporate different types of constraints, even non-additive. (2) Build high accuracy models for the target protein. Homology modeling based on a given sequence alignment between a target sequence and the sequence of a template protein whose structure is known usually involve two steps: loop building, to fix regions of insertion and deletion in the alignment, and side-chain modeling. We will elaborate from our extensive experience in developing solutions to both problems to propose new approaches that circumvent these difficulties. (3) Improve initial models to generate crystal structure quality models using structure refinement. We will follow three directions (a) minimization and molecular dynamics with improved force fields, including quantum mechanical terms, (b) minimization and molecular dynamics with improved implicit solvent models and (c) energy minimization with cooperative many-body energy terms. (4) Organize all computer programs developed within this proposal into a user-centric package, including visual computing tools, to make it accessible to the biological community at large.