Acta Chimica Sinica ›› 2013, Vol. 71 ›› Issue (04): 670-678.DOI: 10.6023/A12121059 Previous Articles    

Article

基于ISC-SVR方法预测Th细胞表位

于畅宇, 刘伟, 刘涛, 宋哲, 朱鸣华   

  1. 大连理工大学高科技研究院 大连 116023
  • 投稿日期:2012-12-17 发布日期:2013-01-31
  • 通讯作者: 刘伟 E-mail:jchjys@dlut.edu.cn

Prediction of Th Cell Epitopes Based on ISC-SVR Method

Yu Changyu, Liu Wei, Liu Tao, Song Zhe, Zhu Minghua   

  1. College of Advanced Science and Technology, Dalian University of Technology, Dalian 116023
  • Received:2012-12-17 Published:2013-01-31

Exogenous antigen proteins are absorbed by antigen presenting cells (APC) and translated into lysosome in which they are degraded into short peptides with different length. A fragment called class II-associated invariant chain peptide (CLIP) lodges in the peptide-binding groove after the Ii chain is progressively degraded by lysosomal proteases. With the help of HLA-DM molecule, CLIP is removed from the binding groove of MHC class II molecule, which makes the degraded peptides enter into the empty binding groove to form steady peptide-MHC class II complexes. The peptide-MHC class II complexes are ultimately trafficked to the cell surface to be recognized by CD4+T cell surface receptor (TCR). This process can activate Th cells to secrete cell factors that promote CTL cells to kill target cells specifically or assist B cells to make antibody against bacteria. Activation of Th cells has important supplementary effect on the cellular immunity and humoral immunity function. In this paper, the support vector regression method combined with iterative self-consistent strategy (ISC-SVR) were used to build the models to predict affinities of exogenous antigen peptides binding to MHC class II molecules. In order to improve the predictive performance, the predictive models were based on 13mer extended core binding sequence rather than traditional 9mer core binding sequence. We computed the affinity of MHC class II molecule binding to peptide with four methods, which were the mean method, the max method, the combine method and the eave method respectively. The predictive performance of the ISC-SVR model is validated on data sets of 17 MHC class II alleles. The results show that the model has certain rationality and accuracy in T cell epitopes prediction. Compared to other predictive models with the same data set, the predictive performance of our model is more satisfying. For each MHC class II allele, the ISC-SVR model got the best AUC value which was used to measure the predictive performance. Furthermore, for HLA DRB1*0101 molecule, we obtained the specificities of MHC class II molecule binding to peptides which were consistent with previous experimental results.

Key words: T cell, epitope, exogenous antigen, MHC class II molecule, support vector regression