含伪结RNA二级结构预测的支持向量机方法  

RNA Secondary Structure Prediction with Plane Pseudoknots Based on Support Vector Machine

作　　者：张洪礼[1] 张娜[1] 王常武[1] 王宝文[1] 刘文远[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北省秦皇岛066004

出　　处：《小型微型计算机系统》2010年第10期1993-1996,共4页Journal of Chinese Computer Systems

基　　金：国家火炬计划项目(2008GH540088)资助;河北省教育厅自然科学研究计划项目(2009339)资助

摘　　要：RNA二级结构预测是生物信息学的一个研究重点和难点.伪结是RNA二级结构中最难预测的一种.本文利用多分类支持向量机对含平面伪结的RNA分子二级结构进行预测.第一步,利用多分类支持向量机进行预测,输出端得到相应碱基的E-NSSEL类别标识.第二步,利用第一步的预测结果,把所有可能配对的标识是否配对通过支持向量机进行判断,并根据判断结果恢复RNA分子二级结构.实验证明,该方法能有效地提高含平面伪结的RNA分子二级结构的预测精度.RNA secondary structure prediction is one of the most important and difficult research areas in Bioinformatics. Pseudoknot prediction is the most difficult research in the area. This paper introduced a new representation of the RNA structure information with plane pseudoknots by the multi-class Support Vector Machine （multi-class SVM ）. First step, a multi-class SVM model was presented to predict RNA secondary structure based on E-NSSEL labels which can express plane pseudoknots. Second step, another SVM was used to judge the two predicted labels whether getting paired or not by considering the labels of neighbor residues. After the two steps, resumed the RNA secondary structure according to the predicted results. Experiment proved that this method could improve the precision of the RNA secondary structure prediction with plane pseudoknots.

关 键 词：多分类支持向量机 RNA二级结够 E-NSSEL标识 平面伪结 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]