DYNAMIC MODELS OF PROTEIN FOLDING

一些调查人员，包括Jernigan和Covell at NIH、加州大学旧金山分校的Dill小组和Shahnowvich的小组在 哈佛大学，曾研究立方晶格的动力学模型 一系列残基的折叠。通常是允许的集合 移动包括摆动(仅限末端)、平面翻转 链所在的两种可能状态之一之间的残留物 进行折弯，曲轴在3-空间中移动。一种能量 矩阵是用来给出接触之间的能量 链上不相邻的残基。大都会 算法用于搜索最低能量状态。什么时候 最低的能量状态被发现链已经折叠 成功了。立方体的理想性质之一 这里描述的格子模型是这样一个事实，即它不是 很难举出长度为27或更长的链的例子 可以用这种方式成功折叠。 虽然立方晶格模型一直是一个有趣的模型 其中研究了折叠问题，并导致了一些 洞察，这并不是一个非常现实的模型 蛋白质。为此，我们进行了一项关于 立方晶格模型，试图理解其 动力学，能量和熵之间的取舍， 目的是在可能的情况下发现一些 可能导致更现实的模型的原则 折叠。有人提出，人们应该努力找到 用于最大化Tf/Tg的模型的参数(布林格尔森， 等人)或Tf/T(Theta)(Camacho&Thirumalai)。中的 这里提到的三个转变温度我们已经发现 只有TF才能帮上忙。TG很难用来获得 足够的精确度是有用的，而T(Theta)则没有 到目前为止，已被证明在预测中是有用的。我们有 开发了一种绘制自由能表面的方法，通过 在Tf上展开一个分子。虽然这不会给出 它提供了自由能表面的完整图像 一张对调整参数非常有用的图片 模型的一部分。通过使用宫泽和杰尼根的能量 矩阵，我们已经能够调整一个 长度为125个残基的序列以获得折叠 对很大一部分试验来说是合理的时间。一个 手稿正在准备中。

A number of investigators, including Jernigan and Covell at NIH, Dill's group at UCSF, and Shahknovich's group at Harvard, have studied cubic lattice models of the dynamic folding of a chain of residues. Usually the set of allowed moves consist of wiggles (ends only), planar flips of a residue between one of two possible states where the chain makes a bend, and crankshaft moves in 3-space. An energy matrix is used to give the energy of contact between residues not adjacent on the chain. The metropolis algorithm is used to search for a lowest energy state. When the lowest energy state is found the chain has folded successfully. One of the desirable properties of the cubic lattice model as here described is the fact that it is not difficult to give examples of chains of length 27 or more that can be successfully folded in this way. While the cubic lattice model has been an interesting model in which to examine the folding problem and has led to some insights, it is not a very realistic model for actual proteins. For this reason we have undertaken a study of the cubic lattice model in an attempt to understand its dynamics, the give and take between energy and entropy, with the purpose of uncovering if possible some general principles that might lead to more realistic models of folding. It has been proposed that one should try to find parameter's for the model which maximize Tf/Tg (Bryngelson, et al.) or Tf/T(theta) (Camacho & Thirumalai). Of the three transition temperatures mentioned here we have found only Tf to be helpful. Tg is difficult to obtain with sufficient accuracy to be useful and T(theta) has not proved to be useful in prediction thus far. We have developed a method of mapping the free energy surface by unfolding a molecule at Tf. While this does not give a complete picture of the free energy surface it does provide a picture that is very useful for adjusting the parameters of the model. By use of the Miyazawa & Jernigan energy matrix we have been able to adjust the parameters of a sequence 125 residues long so as to obtain folding in reasonable times for a large fraction of the trials. A manuscript is in preparation.