Vendruscolo M, Domany E
Department of Physics of Complex Systems, Weizmann Institute of ScienceRehovot, 76100, Israel.
Fold Des. 1998;3(5):329-36. doi: 10.1016/S1359-0278(98)00045-5.
Two problems are of major importance in protein fold prediction: how to generate plausible conformations, and how to choose an energy function to identify the native state. Contact maps are a simple representation of protein structure and offer a promising framework to address these two issues.
In this work we develop Monte Carlo dynamics in contact map space. The procedure is divided into four steps: non-local dynamics, in which large-scale "cluster" moves are performed (clusters are in approximate correspondence with secondary structure elements); local dynamics, in which secondary structure location is optimized; reconstruction, in which the physicality of the contact map is restored; and refinement, which consists of a further Monte Carlo energy minimization in real space. We demonstrate that such a dynamical procedure is effective in producing uncorrelated low-energy states.
The procedure introduced in this paper very effectively generates a representative ensemble of conformations. We are able to show that existing sets of pairwise contact energy parameters are not suitable to single out the native state within this ensemble. The remaining outstanding issue in protein folding is to find an energy function that can discriminate the native state from decoys.
在蛋白质折叠预测中，有两个问题至关重要：如何生成合理的构象，以及如何选择能量函数来识别天然状态。接触图是蛋白质结构的一种简单表示形式，为解决这两个问题提供了一个有前景的框架。
在这项工作中，我们在接触图空间中开发了蒙特卡罗动力学。该过程分为四个步骤：非局部动力学，其中进行大规模的“簇”移动（簇大致对应于二级结构元件）；局部动力学，其中优化二级结构位置；重建，其中恢复接触图的物理性质；以及细化，它包括在实空间中进一步的蒙特卡罗能量最小化。我们证明这种动力学过程在产生不相关的低能状态方面是有效的。
本文介绍的过程非常有效地生成了具有代表性的构象集合。我们能够表明，现有的成对接触能量参数集不适用于在这个集合中挑选出天然状态。蛋白质折叠中剩下的突出问题是找到一种能量函数，能够将天然状态与诱饵区分开来。
