Prediction of Protein Tertiary Structures by Monte Carlo Simulated Annealing and Multicanonical Algorithms

我们一直提倡使用蒙特卡罗模拟退火法和多正则算法来解决蛋白质折叠问题。虽然我们已经在气相中对算法进行了模拟测试，但众所周知，溶剂的影响对于蛋白质结构的形成和稳定性是非常重要的。因此，开发一种在模拟中包含溶剂效应的有效方法是非常重要的。我们在三个不同的模型中测试了包含溶剂效应的情况。第一种是基于与溶剂可及表面积成正比的能量项。第二种是基于标度粒子理论表示疏水效应的能量项。对于第一种可能性，我们比较了5种常用的甲硫氨酸脑啡肽和蛋白质G的溶剂可及表面积参数集的有效性，结果表明其中两种参数是不合适的。对于基于标度粒子理论的第二种可能性，我们对甲硫氨酸脑啡肽进行了蒙特卡罗模拟退火法模拟。我们发现，得到的最低能量构象比气相模拟得到的构象范围更广。这与核磁共振实验一致，表明甲硫氨酸脑啡肽在水溶液中得到了充分的扩展。最后，对于RISM理论的情况，我们计算了我们准备进行比较的六种构象的溶剂化能。我们发现，在这六种结构中，完全伸展的结构的总能量最低。现在，我们准备开始使用蒙特卡罗模拟退火法和/或多正则化算法进行认真的模拟。

We have been advocating the use of Monte Carlo simulated annealing and multicanonical algorithms in the protein folding problem. Although we have been testing the algorithms by simulations in gas phase, it is well known that the effects of solvent are very important for the protein structure formation and stability. It is thus very important to develop an efficient method of inclusion of solvent effects in the simulations. We have tested the inclusion of solvent effects in three different models. First one is based on an energy term that is proportional to solvent accessible surface area. The second one is based on an energy term that represents hydrophobic effects by the scaled particle theory. The third one is based on an energy term that represents all the solvent effects by the RISM theory.For the first possiblity, we compared the validity of 5 commonly used parameter sets for the solvent accessible surface area for Met-enkephalin and Protein G.We have shown that two out of the five are inappropriate. For the second possibility that is based on the scaled particle theory, we performed Monte Carlo simulated annealing simulations for Met-enkephalin. We found that the lowest-energy conformation obtained is more extended than that obtained in gas-phase simulations. This is consistent with NMR experiments, Which indicates that Met-enkephalin is fully extended in aqueous solution. Finally, for the case of RISM theory, we calculated the solvation energies of six conformations that we prepared for comparison. We found that a fully extended structure has the lowest total energy among the six. We are now ready for starting serious simulations by Monte Carlo simulated annealing and/or multicanonical algorithms.