Theoretical Studies of Protein Folding

这个项目的智力目标主要不是预测蛋白质结构，而是了解残基之间的相互作用如何决定球形蛋白质的三维结构（蛋白质折叠问题）。对于这样的理解，使用从头算方法，即仅基于势能函数（包括溶剂的作用）的全局优化，而不使用二级结构预测，同源性建模，线程等。这种方法既需要可靠的势函数，又需要有效的全局优化过程；一个人不能在没有另一个的情况下实现一个。因此，具体目标是(i)改进势函数，（ii）改进全局优化程序，以及（iii）首先将它们应用于已知结构的球状蛋白，然后应用于casp样盲测中未知结构的蛋白质。势函数是通过从头算来改进的，再加上用同样的全局优化程序来改进，这一过程曾用于从头算预测晶体结构（类似于蛋白质折叠问题）。多重极小问题（由于多维势能面存在许多局部极小值）正在被一种新开发的分层方法所克服，该方法包括（除其他程序外）多肽链的联合残差（UNRES）描述、构象空间退火（CSA）方法、蒙特卡罗加能量最小化（MCM）方法及其后代[构象族蒙特卡罗（CFMC）方法]；这种全局优化的分层方法在CASP3盲测试中非常成功，并且在以后类似casp的盲测试中得到了显著改进。有了PI自己的PC-Linux集群，在Cornell理论中心的NT集群和San Diego超级计算机中心的超级计算机上分配了相当多的时间，就可以获得执行该项目所需资源的足够访问权限。提出的理论方法将提供对蛋白质构象和折叠特性的基本理解，并将为进行这项研究的两位研究人员，以及与这些研究人员一起工作的几位博士后和研究生提供培训。

Sheraga, Harold A.MCB-0003722 The intellectual goal of this project is not, primarily, to predict protein structure but, rather, to gain an understanding of how inter-residue interactions determine the three-dimensional structure of a globular protein (the protein folding problem). For such an understanding, use is made of an ab initio approach, i.e. one based solely on the global optimization of a potential energy function (including the role of the solvent), without the use of secondary-structure predictions, homology modeling, threading, etc. Such an approach requires both a reliable potential function and an efficient procedure for global optimization; one cannot implement one without the other. Therefore, the specific aims are (i) to improve the potential function, (ii) to improve procedures for global optimization, and (iii) to apply them, first, to globular proteins of known structure and, then, to proteins of unknown structure in the CASP-like blind tests. The potential function is being improved by ab initio calculations plus refinement by the same global optimization procedure that was used for ab initio predictions of crystal structures (an analog of the protein folding problem). The multiple-minima problem (due to the existence of numerous local minima in the multidimensional potential energy surface) is being surmounted by a newly-developed hierarchical method, which incorporates (among other procedures) a united-residue (UNRES) description of the polypeptide chain, a Conformational Space Annealing (CSA) method, and the Monte-Carlo-plus-energy minimization (MCM) method and its descendant [the Conformation-Family Monte Carlo (CFMC) method]; this hierarchical method for global optimization was very successful in the CASP3 blind test, and is being significantly improved for later CASP-like blind tests. With the PI's own PC-Linux cluster, and considerable allocated time on both the NT clusters of the Cornell Theory Center and the supercomputer at the San Diego Supercomputer Center, sufficient access is available to the necessary resources to carry out this project. The proposed theoretical approach will provide a basic understanding of the conformational and folding properties of proteins, and will provide training, not only for the two research associates carrying out this research, but also for several postdocs and graduate students working together with these research associates.