中文摘要：

目的建立EB病毒（EBV）抗原表位的理论计算模型并用于其定量预测，为肿瘤免疫的多肽疫苗设计提供理论基础。方法从表位数据库中收集33条EBV抗原表位，使用逐步回归（STR）方法筛选2个对平衡解离常数（KD）贡献较大的结构参数用于表位的结构表征，最后用多元线性回归（MLR）方法建立结构参数与KD的定量构效关系（QSAR）模型。结果该模型具有较好的稳定性（R^2=0．637，Q^2=0．581）与预测能力（R^2test=0．501）。结论此方法可确定表位中各个氨基酸的物理性质对于平衡解离常数的贡献，为表位设计与改造提供直接线索；STR和MLR相结合建模方法具有物化意义明确、易于解释及操作简便易行等优点。

英文摘要：

Epstein-Barr virus （-EBV） is highly associated with several neoplastic diseases. The development of vaccine based on the antigenic epitope is very important for immunotherapies of human cancers. It is necessary to construct theoretical model for quantitative prediction of epitope, because the identification and screen of epitope need long time and high cost through experiments. In this study, 33 epitopes from EBV had been collected from epitope database. Then, the properties with significant contribution for equilibrium dissociation constant （KD） were screened by the stepwise, which were used to characterize the epitopes. Finally, the quantitative structure-activity relationship model between structural variables and KD was constructed by multiple linear regressions （MLR）. The QSAR model has good reliability （R^2=0.637, Q2=0.581） and predictive ability （R^2test=0.501）, and can provide more clues for the design and modification of epitope. The QSAR model constructed by STR and MLR has some advantages, such as good reliability and predictive ability, definitive physiochemical meaning and easier operation, and it can guide the rational design and structural modification of epitope directly.