基于茎区的神经网络方法预测RNA二级结构  

Prediction of RNA Secondary Structure by Using the Neural Network of Stems

作　　者：许丹[1] 王爱荣[1] 李金铭[1] 

机构地区：[1]福建农林大学计算机与信息学院,福建福州350002

出　　处：《广西师范大学学报（自然科学版）》2011年第3期142-146,共5页Journal of Guangxi Normal University:Natural Science Edition

基　　金：福建省自然科学基金资助项目(2006J0018);国家自然科学基金资助项目(30800713)

摘　　要：RNA二级结构预测是生物信息学的一个重要研究内容。作为预测方法之一的神经网络已被广泛应用于蛋白质结构预测,但在RNA二级结构的应用甚少。本文改进传统预测RNA二级结构的Hopfield神经网络。算法以茎作为网络神经元,通过与相似结构茎区的比对,初始化神经元,并据此修改网络的激励系数。实验把改进后算法与改进前2种算法、Mfold、RNAStructure比较,结果表明本文提出的算法对序列长度较小并且保守性较好的tRNA分子有很好的效果。RNA secondary structure prediction is an important research field in bioinformatics.As one of the forecasting method,neural network has been widely used in protein structure prediction but very little in RNA secondary structure.The traditional prediction of RNA secondary structure using Hopfield neural network is improved in this paper.Stem is used as the neuron of network in the algorithm.The network incentive factor and the initial value of neurons are modified by alignment with the stem area of similar structure.The improved algorithm is compared to two kinds of unimproved algorithms,Mfold and RNAStructure.Experiments shows that the proposed algorithm has very good results in smaller and better conservative tRNA molecules.

关 键 词：HOPFIELD神经网络 RNA二级结构 茎区 TRNA 

分 类 号：Q811[生物学—生物工程] TP389.1[自动化与计算机技术—计算机系统结构]