论文部分内容阅读
从低同源关系的氨基酸序列预测蛋白质的三维结构被称为从头预测,它是计算生物学领域中的挑战之一.蛋白质骨架预测是从头预测的必要先导步骤.本文应用一种基于共享信息素的并行蚁群优化算法,在现有能量函数指导下,通过不同能量项之间的定性互补,构建具有最低能量的蛋白质骨架结构,并通过聚类选择构象候选集合中具有最低自由能的构象.在CASP8/9所公布的从头建模目标上应用了该方法,CASP8的13个从头建模目标中,模型1中有2个目标的预测结果超过CASP8中最好的结果,7个位列前10名;CASP9的29个从头建模目标中,候选集中的最佳结果中有20个进入Server组的前10名,模型1中有11个进入前10名.本文的结果说明融合多个不同的能量函数指导并行搜索,可以更好地模拟天然蛋白质的折叠行为.同时,在本算法载体上实现了不同种类搜索策略的融合并行,对于用非确定性算法解决类似的优化问题来说也是一种新颖的方法.
It is one of the challenges in the field of computational biology to predict the three-dimensional structure of a protein from a sequence of amino acids with low homology, which is one of the challenges in the field of computational biology.Protein skeleton prediction is an essential leading step in ab initio prediction.In this paper, Under the guidance of the existing energy function, a parallel ant colony optimization algorithm is used to construct the protein skeleton structure with the lowest energy through the qualitative complementation between different energy terms. The conformation with the lowest free energy in the candidate conformation set is selected through clustering. Applying this approach to the de novo modeling goals announced by CASP8 / 9, out of the 13 de novo modeling goals for CASP8, two in model 1 predicted better than the best in CASP8 with seven 10 of the 29 modeling goals of CASP9, 20 of the best results in the candidate set entered the top 10 of the Server group, and 11 of the models entered the top 10. The results of this paper show that the fusion of a number of different Of the energy function to guide the parallel search, you can better simulate the natural protein folding behavior.At the same time, in this algorithm carrier to achieve different kinds of search strategy fusion parallel, for non-deterministic calculation Similar to solve optimization problems is also a novel approach.