Standley D M, Gunn J R, Friesner R A, McDermott A E
Department of Chemistry and Center for Biomolecular Simulation, Columbia University, New York, NY 10027, USA.
Proteins. 1998 Nov 1;33(2):240-52.
We describe an improved algorithm for protein structure prediction, assuming that the location of secondary structural elements is known, with particular focus on prediction for proteins containing beta-strands. Hydrogen bonding terms are incorporated into the potential function, supplementing our previously developed residue-residue potential which is based on a combination of database statistics and an excluded volume term. Two small mixed alpha/beta proteins, 1-CTF and BPTI, are studied. In order to obtain native-like structures, it is necessary to allow the beta-strands in BPTI to distort substantially from an ideal geometry, and an automated algorithm to carry this out efficiently is presented. Simulated annealing Monte Carlo methods, which contain a genetic algorithm component as well, are used to produce an ensemble of low-energy structures. For both proteins, a cluster of structures with low RMS deviation from the native structure is generated and the energetic ranking of this cluster is in the top 2 or 3 clusters obtained from simulations. These results are encouraging with regard to the possibility of constructing a robust procedure for tertiary folding which is applicable to beta-strand containing proteins.
我们描述了一种改进的蛋白质结构预测算法，假设二级结构元件的位置已知，特别关注含β链蛋白质的预测。氢键项被纳入势函数，补充了我们之前基于数据库统计和排除体积项组合开发的残基-残基势。研究了两种小型混合α/β蛋白，1-CTF和BPTI。为了获得类似天然的结构，有必要允许BPTI中的β链从理想几何形状大幅扭曲，并提出了一种有效实现此目的的自动化算法。包含遗传算法组件的模拟退火蒙特卡罗方法用于生成一组低能结构。对于这两种蛋白质，都生成了一组与天然结构具有低均方根偏差的结构，并且该组结构的能量排名在模拟得到的前2或3组中。这些结果对于构建适用于含β链蛋白质的稳健三级折叠程序的可能性而言是令人鼓舞的。
