点击次数：

发表时间：2014-01-01

发表刊物：大连理工大学学报

期号：1

页面范围：28-36

ISSN号：1000-8608

摘要：In the design of peptide-based or other defined antigen-based vaccines ,it is important to know w hich fragments of pathogen-derived proteins would bind to the M HC Ⅱ molecules .Most studies of the M HC Ⅱ epitope prediction rarely gave the quantitative analyses of binding specificities .So the accuracy of these models still needs to be improved .AUC Optimized Gibbs (AOG) method uses the homology reduced AUC , rather than the relative entropy to guide the sampler . It makes both the positive and negative information of the samples be incorporated into the model . AOG achieves average AUC values of 0 .771 and 0 .713 on the ten original and homology reduced HLA-DR4 (B1 * 0401) epitope benchmarks ,which are better than 0 .744 and 0 .673 by the Gibbs sampling method . In the quantitative IEDB M HC-Ⅱ benchmarks , AOG achieves an average AUC value of 0 .766 , compared to 0 .718 by the TEPITOPE .A detailed inspection of information extracted from HLA-DR4 (B1 * 0401 ) data allows the identification of positions with obvious specificities ,i .e .,P1 ,P4 ,P6 and P9 positions ,which have distinct influence on the M HC-peptide binding .

备注：新增回溯数据