基于神经网络集成的蛋白质二级结构预测模型研究  

Study on the Prediction Model Based on Neural Network Ensemble of Protein Secondary Structure

作　　者：刘军[1,2] 马文丽[1,3] 姚文娟[1] 郑文岭[1,3] 

机构地区：[1]上海大学电子生物技术研究中心,上海200072 [2]江苏财经职业技术学院电子系,江苏淮安223003 [3]南方医科大学分子生物学研究所,广东广州510515

出　　处：《安徽农业科学》2009年第27期12884-12886,共3页Journal of Anhui Agricultural Sciences

摘　　要：[目的]探讨基于CPN神经网络集成的蛋白质二级结构预测模型的效果。[方法]借助神经网络集成方法对从36个蛋白质提取的共4 000个氨基酸进行预测研究,其数据集是从HSSP数据库中提取的数据经过处理后得到的评测数据库,同时在Profile编码中引进了CPN网络算法的概念。[结果]基于CNP网络的神经网络集成预测模型可以取得很好的预测结果,把蛋白质二级结构预测的平均精度提高了17.74%。同时,所用的Profile编码和CPN网络算法在很大程度上为系统模型引入较多的生物信息和联系,而这一点对蛋白质二级结构预测非常重要。[结论]该研究为蛋白质二级结构预测准确率的提高奠定了基础。[ Objective ] The aim was to discuss the effect of the protein secondary structure prediction model based on CPN neural network ensemble. [ Method] Total 4 000 amino acids extracted from 36 proteins were predicted by the method of neural network ensemble. The data set was an evaluation database and it was obtained after the data extracted from HSSP database were treated and meanwhile the concept of CPN network algorithm was introduced during Profile coding. [ Result] The prediction model of neural network ensemble based on CPN network could obtain good prediction result and the average precision of protein secondary structure prediction could be increased by 17.74%. Meanwhile, the used Profile coding and CPN network algorithm introduced into more biological information and relation for system model to a great extent, which was important to protein secondary structure prediction. [ Conclusion ] The research laid the foundation for the advance of prediction accuracy of protein secondary structure.

关 键 词：蛋白质二级结构 CPN 神经网络集成 

分 类 号：Q581.1[生物学—生物化学]