中文摘要：

目的主要组织相容复合物（MHC）在细胞免疫中具有重要作用,MHC与表位亲和力的大小直接影响T细胞免疫响应的强弱,因此准确预测MHC-表位的结合亲和力可以提高表位的筛选效率并且降低实验成本。方法筛选89种天然氨基酸的理化性质用于15个MHCⅠ类分子亚型表位的结构表征,然后使用逐步线性回归（STR）方法优选对亲和力具有重要贡献结构参数建立预测模型,最终用多元线形回归法（MLR）方法建立了15个MHC亚型的QSAR模型。结果建立的QSAR模型具有稳定性高、预测能力强等优点,能确定表位中各氨基酸及其相应理化性质队对亲和力的贡献,并可很好解释MHC与表位的相互作用机理。结论 STR-MLR建模方法具有计算简便、易于解释、性能优异、物化意义明确等优势。

英文摘要：

Major histocompatibility complex（MHC） plays a pivotal role in T cellar immunity,and the T cell immune responses rely on the binding affinity between MHC and epitope.Therefore,the accurate prediction of MHC-epitope binding affinity could greatly expedite epitope screening and reduce cost and experimental efforts.In this study,89 physicochemical properties were screened to characterize the epitopes from 15 class I MHC subtypes,then the important variables were optimized for quantitative structure-activity relationship（QSAR） modeling by stepwise regression,finally 15 QSAR models were constructed by multiple linear regression.These QSAR models have good stability and predictive ability,which could determine the contributions of position of epitope and corresponding physicochemical properties to the binding affinity,and express the interactive mechanism between MHC and epitope.Meanwhile,the STR-MLR had many advantages,such as good performance,definite physiochemical indication,and easy to calculate and explain.