基于交叉覆盖算法的蛋白质二级结构预测方法  

A Method for Protein Secondary Structure Prediction Based on Alternative Covering Algorithm

作　　者：张燕平[1] 章晶[1] 徐庆鹏[1] 朱远枫[1] ZHANG Yan-ping, ZHANG Jing, XU Qing peng, ZHU Yuan-feng （Key Lab. of Intelligent Computing ＆ Signal Processing, Ministry of Education, Anhui University, Hefei 30039, China）

机构地区：[1]安徽大学计算智能与信号处理教育部重点实验室,安徽合肥230039

出　　处：《电脑知识与技术》2009年第1期206-207,共2页Computer Knowledge and Technology

基　　金：国家重点基础研究973计划项目（2004CB318108,2007CB311003）,国家自然科学基金项目（60675031）,安徽大学211工程学术创新团队安徽大学人才队伍建设经费资助

摘　　要：蛋白质二级结构预测在蛋白质空间结构预测中起着承上启下的重要作用。近年来，大量的方法应用于二级结构预测中，其中，神经网络算法效果较好。但是，由于传统的神经网络存在结构复杂、学习速度慢、运行效率低、处理海量数据困难的缺陷，大大影响了预测的效果，因此，该文将一种基于构造性神经网络算法，也就是交叉覆盖算法应用于蛋白质二级结构预测中，另外，为了引入更多的同源家族结构的信息，采用了基于概率的Profile编码方式。通过实验证明将交叉覆盖算法运用在蛋白质二级结构预测中的可行性．并且比传统的神经网络方法有了更高的准确率。Prediction of protein secondary structure play an important role in protein structure forecast,which is the connection of protein secondary structure and protein tertiary structure.P, ecently,a lot of methods were applied to the prediction of protein secondary structure,among these methods,neural networks have higher precision than others.But traditional neural network have some disadvantages such as complex structure,low speed of learuing,low efficiency and it has difficult in dealing with huge datas.Therefore,a constructive neural networks,namely alternative covering algorithm was applied to the prediction of protein secondary structure in this paper.Moreover, this paper used Profile encoding for protein sequence to incorporate more information of homologous family structure. Results in the present work show that the proposed method is feasible and more effective in comparison to traditional neural network.

关 键 词：交叉覆盖算法 蛋白质二级结构 Profile编码 

分 类 号：TP183[自动化与计算机技术—控制科学与工程;自动化与计算机技术—控制理论与控制工程]