使用支持向量机的方法预测膜蛋白的类型  

PREDICTING MEMBRANE PROTEIN TYPES BY USING SUPPORT VECTOR MACHINE

作　　者：王婷[1,2] 胡秀珍[1] 

机构地区：[1]内蒙古工业大学理学院,呼和浩特010051 [2]长治职业技术学院,山西长治046000

出　　处：《内蒙古工业大学学报（自然科学版）》2010年第4期241-246,共6页Journal of Inner Mongolia University of Technology：Natural Science Edition

基　　金：国家自然科学基金资助项目(30960090);内蒙古自治区高等学校科学研究项目(NJZY08059)

摘　　要：根据膜蛋白的特点对特定的氨基酸片段进行打分,将打分值与离散增量值共同作为序列的信息参数,使用支持向量机的方法对膜蛋白的类型进行预测,获得了非常好的预测结果。对8类膜蛋白的Jack knife检验结果达到了91.81%,比Chou等人使用相同数据集的结果提高了近7个百分点。另外,使用同样的方法对6类膜蛋白的类型也作了预测,结果再次表明该方法对于膜蛋白类型的预测是非常有效的。By using the composite vector of scoring function with increment of diversity to express the information of sequence, a support vector machine （SVM） algorithm for predicting eight types of membrane proteins is proposed. The overall jackknife success rate reaches 91.81%, which is almost seven percent higher than Chou＇s results with use of the same database. In order to evaluate the predictive method, six more types of membrane proteins are also predicted by the method presented in this paper. Better results are obtained again which proves that this method is indeed very effective for predicting membrane protein types.

关 键 词：膜蛋白 矩阵打分 离散增量 支持向量机 

分 类 号：Q61[生物学—生物物理学]