Predicting side-chain packings in proteins using evolution strategies

9813449格林伍德进化计算（EC）技术在分子生物学中的各种优化问题的应用已经变得越来越普遍。 相当感兴趣的是预测蛋白质中的侧链构象[1，2]。 由于蛋白质内部侧链的高密度，侧链构象由包装考虑决定，这涉及与其他侧链的空间效应。 解决方案是困难的，因为有大量的组合可能。 然而，这个问题的有效解决方案是必不可少的，大多数从头计算蛋白质结构预测程序。 它也将有助于确定允许的突变位点，以增加蛋白质的稳定性和大规模的蛋白质重新设计。优化问题的搜索空间中的每个点都代表一个唯一的解决方案;因此，优化算法是搜索算法。 EC算法是概率优化算法，其使用达尔文进化原理进行搜索。 对于侧链堆积问题（SCPP），群体中的每个个体代表不同的侧链构象。 在每一代（EC算法的迭代），随机再生算子扰动现有的构象，以产生新的构象。 优化问题提供了一个衡量解决方案的质量，这是由个人的适应度。(In SCPP高适合度可以表示最小的货车德瓦尔斯相互作用能。高度健康的个体将倾向于在后代中生存和繁殖。 研究表明，在分子生物学优化问题中，EC算法比Monte Carlo方法优越上级[2，3，4]。遗传算法（GA）已成为应用于SCPP的主要EC算法（例如，见[5]）。 然而，CA的某些方面表明它们可能不是分子生物学问题的最佳EC算法[6，7]。 这项工作将使用ES找到最佳的包装安排的侧链。ES算法从未被用于寻找蛋白质结构预测问题的解决方案。 尽管如此，如果先前的原子簇工作的结果是任何指示的话，ES算法很可能被证明是优于GA方法的上级。对于这项工作，蛋白质骨架的Cc'位置将被假定为固定的，但是骨架的肽基团将被假定为可旋转的，并且它们的最佳取向将被发现作为侧链包装优化问题的一部分。 键长和键角也是固定的，因此EC必须仅识别肽取向角和侧链扭转角。 将使用包括货车德瓦尔斯相互作用、氢键和疏水效应的能量函数来确定适应度。*

9813449GreenwoodThe application of evolutionary computation (EC) techniques to various optimization problems in molecular biology has become increasingly common. Of considerable interest is the prediction of side-chain conformations in proteins [1, 2]. Because of the high density of side-chains in the protein interior, side-chain conformations are determined by packing considerations, which involve steric effects with other side-chains. The solution is difficult because of the large number of combinations possible. Nevertheless, an efficient solution to this problem is essential for most ab initio protein structure prediction procedures. It will also aid in determining permissible mutation sites for increased protein stability and for a large scale protein re-design.Each point in the search space of an optimization problem represents a unique solution; optimization algorithms are therefore search algorithms. EC algorithms are probabilistic optimization algorithms which conduct searchs using the principles of Darwinian evolution. For the side-chain packing problem (SCPP), each individual in a population represents a distinct side-chain conformation. During each generation (iteration of the EC algorithm), stochastic reproduction operators perturb existing conformations to produce new conformations. The optimization problem provides a measure of the solution's quality which is indicated by the individual's fitness. (In SCPP high fitness may represent a minimal van der Waals interaction energy.) Highly fit individuals will tend to survive and reproduce in future generations. Studies have shown that the EC algorithms are superior to Monte Carlo methods in molecular biology optimization problems [2, 3, 4].Genetic Algorithms (GAs) have been the predominant EC algorithm applied to SCPP (e.g., see [5]). Yet, there are certain aspects of CAs which suggests they may not be the best EC algorithm for molecular biology problems [6, 7). This work will use an ES to find optimum packing arrangements for side-chains.' The ES algorithm has never been used to find solutions to protein structure prediction problems. Nevertheless, the ES algorithm may well prove to be superior to the GA approach if the results of previous work with atomic clusters are any indication [7].For this work the Cc' positions of the protein backbone will be assumed fixed, but the peptide groups of the backbone will be assumed rotatable and their optimum orientations will be found as a part of the side-chain packing optimization problem. The bond lengths and angles will also be fixed so the EC must only identify the peptide orientation angles and the side-chain torsion angles. The fitness will be determined using an energy function that includes the van der Waals interaction, hydrogen bonding, and hydrophobic effects.***