化学学报 ›› 2010, Vol. 68 ›› Issue (18): 1821-1828. 上一篇    下一篇

研究论文

应用独立成分分析方法预测CTL表位

董素梅1,2,宋哲1,3,刘涛1,2,朱鸣华1,刘伟*,1,2   

  1. (1大连理工大学高科技研究院 大连 116023)
    (2大连理工大学物理与光电工程学院 大连 116023)
    (3大连交通大学理学院 大连 116028)
  • 投稿日期:2009-12-25 修回日期:2010-03-19 发布日期:2010-05-20
  • 通讯作者: 刘伟 E-mail:jchjys@dlut.edu.cn;

Prediction of the CTL Epitope Based on Independent Component Analysis Method

Dong Sumei1,2 Song Zhe1,3 Liu Tao1,2 Zhu Minghua1 Liu Wei*,1,2   

  1. (1 College of Advanced Science and Technology, Dalian University of Technology, Dalian 116023)
    (2 Department of Physics, Dalian University of Technology, Dalian 116023)
    (3 School of Science, Dalian Jiaotong University, Dalian 116028)
  • Received:2009-12-25 Revised:2010-03-19 Published:2010-05-20
  • Contact: 刘伟 LIU E-mail:jchjys@dlut.edu.cn;

基于独立成分分析方法分别采用3 z-scale和5 z-scale氨基酸结构描述符, 建立了抗原肽与MHC分子(major histocompatibility complex, MHC)相互作用结合的定量构效关系模型. 该两个模型训练集样本数是316, 预测集样本数是786. 结果表明: 3 z-scale模型的预测准确度和AUC值分别为70.3%, 0.70; 5 z-scale模型的预测准确度和AUC值分别为70.9%, 0.79. 本文建立CTL表位预测模型对进一步了解抗原肽与MHC I类分子相互作用机理具有一定的帮助.

关键词: CTL表位, 独立成分分析, 氨基酸描述符, 定量构效关系

The quantitative structure-activity relationship (QSAR) models of peptide binding to MHC molecule are studied through the independent component analysis (ICA) method. The three z amino acid descriptors and the five z amino acid descriptors are used respectively in the models. A training set of 316 peptides and a test set of 786 peptides are used. The predictive accuracy and the area under the receiver operator characteristics curve (AUC) of the three z amino acid descriptors model are 70.3% and 0.70 respectively. The predictive accuracy and AUC of the five z amino acid descriptors are 70.9% and 0.79 respectively. This study helps in further understanding the mechanism of the interaction between MHC molecules and peptides.

Key words: cytotoxic T lymphocyte epitope, independent component analysis, amino acid descriptor, QSAR