摘要
目的 蛋白转导域（PTD）是一类能携带分子穿越细胞膜的短肽,利用支持向量机对多肽片段PTD进行预测.方法 对来源于SwissProt数据库的多肽序列用68个特征值描述其整体和局部的理化特性以及空间结构特征,利用支持向量机（SVM）和直推式支持向量机（TSVM）并结合聚类的方法进行PTD的预测.结果 5次交叉验证的结果显示,TSVM的预测准确率达到（94±4）%,SVM预测准确率达到（94±5）%.2种预测方法共同预测了1210个可能的PTD片段.结论 TSVM和SVM均显示了很好的预测性能,预测的PTD为实验方法有目的 地发现、确认PT提供了基础.
Objective To predict protein transduction domain （PTD） which is a short peptide with the ability to pull diverse molecules across cell membranes.Methods Every peptide segment including PTDs and peptide sequence from SwissProt database was represented by 68 numerical values,which reflected their physicochemical and conformational properties related to the PTD＇ s membrane penetrating function.Transductive support vector machine （TSVM） and support vector machine（SVM）,combined with cluster method,was introduced to predict new PTDs from the peptide segments,which were extracted from SwissProt database.Results TSVM prediction model achieved 94%±4% accuracy and SVM model achieved 94%±5% accuracy.1210 possible PTDs were predicted using the classifiers based on these two models.Conclusion The research provides a guide to find more PTDs in molecular biology experiments and will be helpful in the understanding of the mechanism of PTDs and their function in vivo.
出处
《国际生物医学工程杂志》
CAS
北大核心
2010年第4期205-208,共4页
International Journal of Biomedical Engineering
关键词
蛋白转导域
支持向量机
预测
生物信息学
Protein transduction domain
Support vector machine
Prediction
Bioinformatics
