Tanaka S, Scheraga H A
Macromolecules. 1977 Mar-Apr;10(2):305-16. doi: 10.1021/ma60056a016.
One-dimensional short-range interaction models for specific-sequence copolymers of amino acids have been developed in this series of papers. In this paper, a general method for predicting protein conformation (that is based on a one-dimensional short-range interaction model, and eliminates the need for the empirical rules introduced in papers III and IV) is described. The present method involves the use of conformational (or conformational-sequence) probabilities of higher order than the first- or second-order probabilities used in papers IV and V, i.e., it treats a sequence of any number of residues; it thus alters the predictive methods that involved empirical rules in papers III and IV, and low-order (first- or second-order) probabilities in papers IV and V. The general method is applied here to the prediction of the backbone conformations of proteins, using the three-state model [helical (h), extended (epilson), and other coil (c) states] proposed in the theoretical formulation of paper II. The statistical weights in the three-state model are evaluated from the atomic coordinates of the x-ray structures of 26 proteins. The conformational-sequence probabilities (taken for three consecutive residues for numerical computation in this paper) are calculated for all possible triads (i.e., for all possible combinations of the three states, h, epilson, and c for each residue) for bovine pancreatic trypsin inhibitor and clostridial flavodoxin, in order to select the most probable conformations of these proteins. The predicted results for these proteins are compared to those predicted in paper III and to those observed experimentally. The method is applied further to the prediction of the backbone structures of homologous neurotoxin proteins whose amino acid sequences are known but whose x-ray structures are not. The effects of variation in the amino acid sequence on the conformations of the backbones are discussed from the point of view of the homologies in the amino acid sequences of 19 neurotoxins. Application of the present general predictive method to a four- and a multistate model is also described.
在这一系列论文中，已经开发出了用于特定序列氨基酸共聚物的一维短程相互作用模型。在本文中，描述了一种预测蛋白质构象的通用方法（该方法基于一维短程相互作用模型，并且无需论文III和IV中引入的经验规则）。本方法涉及使用比论文IV和V中使用的一阶或二阶概率更高阶的构象（或构象-序列）概率，即它处理任意数量残基的序列；因此，它改变了论文III和IV中涉及经验规则以及论文IV和V中低阶（一阶或二阶）概率的预测方法。这里将该通用方法应用于蛋白质主链构象的预测，使用论文II理论公式中提出的三态模型[螺旋（h）、伸展（ε）和其他卷曲（c）状态]。三态模型中的统计权重是根据26种蛋白质的X射线结构的原子坐标评估的。为了选择这些蛋白质最可能的构象，计算了牛胰蛋白酶抑制剂和梭菌黄素氧还蛋白所有可能三联体（即每个残基的h、ε和c三种状态的所有可能组合）的构象-序列概率（本文中为连续三个残基进行数值计算）。将这些蛋白质的预测结果与论文III中的预测结果以及实验观察结果进行比较。该方法进一步应用于预测氨基酸序列已知但X射线结构未知的同源神经毒素蛋白质的主链结构。从19种神经毒素氨基酸序列的同源性角度讨论了氨基酸序列变化对主链构象的影响。还描述了本通用预测方法在四态和多态模型中的应用。
