Internal Bonding in Proteins

我们的长期目标是确定蛋白质构象如何在各种疾病中发挥作用。我们的特定 目的是计算蛋白质、蛋白质-蛋白质和蛋白质-核酸复合体的三维结构，以及 这些折叠路径通向这些实体，使用了一种基于物理层次的方法。广泛性 构象搜索是用我们的联合残基(UNRES)力场进行的，其中一个多肽 链被表示为虚键C(α)-C(α)和C(α)-SC链，并且所得到的 构象被转换为全原子表示，并在全原子水平上进行精炼。UNRES有 作为联合剩余链的限制自由能函数，其平均度数为 传递到虚键几何时失去的自由。久保的集群累积量理论已经被 用来推导相应自由能项的解析表达式，这使我们能够表达 多体解析项，对于复制规则的α-螺旋和β-折叠是必不可少的 结构。我们的方法在最近的两个盲蛋白质结构预测测试中获得了成功：CASP5和 我们预测了四个靶的完整结构，其中两个是α螺旋的，两个是 阿尔法/贝塔蛋白，以及其他目标的大片段。我们从预测折叠路径开始，通过 为UNRES力场实施朗之万动力学。使用UNRES/MD，我们折叠了一个75-残基 蛋白质从延长的构象到天然结构在单个处理器上大约5小时；这 这意味着，与全原子MD相比，UNRES/MD的时间表增加了10,000倍，是一个 研究可达毫秒级的蛋白质折叠途径的实用方法。UNRES/MD 然而，该方法在目前的参数化方法中忽略了位形熵。我们 将通过修改UNRES的参数来纠正这一缺陷。我们还将进一步发展：(A)我们的 用较短的MD运行取代局部极小化的全局优化算法，从而将 构型熵，(B)我们的全原子力场，和(C)一种将UNRES折叠路径转换为 全原子路径。我们还将扩展UNRES方法，以派生一支基于联合残基物理的部队 研究蛋白质-核酸相互作用的领域。

Our long-term objective is to identify how protein conformation plays a role in various diseases. Our specific aims are to compute the 3D structures of proteins, protein-protein and protein-nucleic acid complexes, and the folding pathways leading to these entities, using a hierarchical physics-based method. Extensive conformational search is performed with our united-residue (UNRES) force field, in which a polypeptide chain is represented as a virtual-bond C(alpha)-C(alpha) and C(alpha)-SC chain, and the resulting conformations are converted to an all-atom representation and refined at the all-atom level. UNRES has been derived as a restricted free energy function of united-residue chains averaged over the degrees of freedom that are lost when passing to the virtual-bond geometry. Kubo's cluster cumulant theory has been used to derive analytical expressions for the respective free-energyterms, which enabled us to express the multibody terms analytically, which are essential for reproducing regular alpha-helical and beta-sheet structures. Our approach was successful in two recent blind protein structure prediction tests: CASP5 and CASP6; we predicted complete structures of four targets of which two were alpha-helical and two were alpha/beta proteins, and large segments of other targets. We made a start on predicting folding pathways by implementing Langevin dynamics for the UNRES force field. Using UNRES/MD, we folded a 75-residue protein from the extended conformation to the native structure in about 5 hours on a single processor; this means that UNRES/MD provides a 10,000 fold increase in the time scale compared to all-atom MD and is a practical method for studying protein-folding pathways up to the millisecond scale. The UNRES/MD approach, however, suffers from the neglect of configurational entropy in the present parameterization. We will rectify this deficiency by revising the parameterization of UNRES. We will also further develop: (a) our global-optimization algorithms by replacing local minimization with short MD runs, thereby incorporating configurational entropy, (b) our all-atom force field, and (c) a method to convert UNRES folding pathways to all-atom pathways. We will also extend the UNRES approach to derive a united-residue physics-based force field to study protein-nucleic acid interactions.