蛋白质构象空间局部增强差分进化搜索方法  

Local Enhancement Differential Evolution Searching Method for Protein Conformational Space

作　　者：董辉[1] 郝小虎[1] 张贵军[1] 

机构地区：[1]浙江工业大学信息工程学院,杭州310023

出　　处：《计算机科学》2015年第B11期22-26,共5页Computer Science

基　　金：国家自然科学基金(61075062;61379020);浙江省自然科学基金(LY13F030008);浙江省科技厅公益项目(2014C33088);浙江省重中之重学科开放基金(20120811);杭州市产学研合作项目(20131631E31)资助

摘　　要：针对蛋白质构象空间搜索问题,提出一种蛋白质构象空间局部增强差分进化搜索方法。在差分进化算法框架下,采用Rosetta Score3粗粒度知识能量模型有效降低构象空间的搜索维数,加快算法收敛速度;引入基于知识的片段组装技术可以有效提高预测精度;利用Monte Carlo算法良好的局部搜索性能对种群做局部增强,以得到更为优良的局部构象;结合差分进化算法较强的全局搜索能力,可以对构象空间进行更为有效的采样。5个测试蛋白实验结果表明,所提算法具有较好的搜索性能和预测精度。A local enhancement differential evolution searching method for protein conformational space was proposed to address the searching problem of protein conformational space. On the framework of differential evolution algorithm, Rosetta Score3 coarse-grained energy model was employed for decreasing the dimension of searching space and improving the convergence rate of algorithm. The knowledge-based fragment assembly technique was introduced for improving the accuracy of prediction. For getting better local near-native conformation, local enhancement operation was done with taking advantage of the well local search performance of Monte Carlo algorithm. The well global searching capacity of the differential evolution algorithm was combined for sampling the whole conformational space effectively. The experiment results on 5 test proteins verify the superior searching performance and prediction accuracy of the proposed method.

关 键 词：蛋白质结构预测 差分进化算法 粗粒度能量模型 片段组装 MONTE Carlo 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]