中文摘要：

目的：采用人工神经网络模型及在线软件BIMAS和SYFPEITHI共同预测汉滩病毒核蛋白HLA—A^＊02限制的CTL表位，并对CTL表位进行实验鉴定。方法：①选取2004-11解放军第四军医大学唐都医院住院的肾综合征出血热患者1例，汉滩病毒特异性IgM抗体为阳性。无菌抽取肾综合征出血热患者外周血，分离外周血单个核细胞，液氮冻存备用。②采用JavaNNS软件模拟神经系统对信息的处理过程，建立人工神经网络模型。训练人工神经网络的9肽包括584种HLA—A^＊02结合肽和130种HLA—A^＊02非结合肽，随机数字表法分为3部分：80％为训练肽（用来训练模型，使其学习其中隐含的规律），10％为确认肽（防止训练过度，使训练在适当的时候终止），10％为试验肽（对训练好的模型进行初步的评价）。无、低、中、高亲和力9肽训练人工神经网络的输出值分别为0．1，0．4，0．6，0．8。因此，人工神经网络模型的默认阈值为0．4。此外，还应用在线软件BIMAS和SYFPEITHI，共同预测汉滩病毒核蛋白的CTL表位。③采用ELISPOT鉴定汉滩病毒核蛋白特异的HLA—A^＊02限制的CTL表位。结果：①人工神经网络、BIMAS和SYFPEITHI预测的准确度：3种方法预测的灵敏度分别达到0．89，0．78，0．97，特异性分别为0．90，0．87，0．20。三者共同预测到23个汉滩病毒核蛋白特异的CTL表位。②HLA—A^＊02限制的CTL表位的实验鉴定：在可刺激肾综合征出血热患者外周血单个核细胞产生阳性反应的3条15肽中，预测到了4个HLA—A ^＊02限制的CTL表位。经ELISPOT实验证实，其中的2种9肽可刺激HLA—A^＊02阳性的肾综合征出血热患者外周血单个核细胞分泌γ干扰素。结论：肾综合征出血热患者中存在着汉滩病毒特异的CTL应答。汉滩病毒核蛋白CTL表位的有效预测，大幅度减少了合成候选9肽的数量，降低了成本。为肾综合

英文摘要：

AIM： To predict HLA-A^＊02 restricted CTL epitopes within nucleocapsid protein of hantaan virus （HTNV-NP） with artificial neural network （ANN） and two online algorithms BIMAS and SYFPEITHI and identify the CTL epitopes experimentally. METHODS： ①One patient with hemorrhagic fever with renal syndrome （HFRS） hospitalized in Tangdu Hospital, Fourth Military Medical University of Chinese PLA was recruited in this study in November 2004. Hantaan virus special IgM antibody was positive. Peripheral blood was isolated from HFRS patients under sterile condition. Peripheral blood mononuclear cells （PBMC） were isolated from the patient and cryopreserved until use. ②ANN was established with JavaNNS software simulating the information-processing procedure of neural system. The training sets which comprised 584 known HLA-A^＊02 binding and 130 non-binding 9-met peptide sequences were divided at random into three parts： 80 percent were used to train the model for learning the implied law, 10 percent to stop the training to prevent excess training, and 10 percent to test the performance of the trained ANN model. Binding scores in ANN training were 0.1, 0.4, 0.6 and 0.8 for non-, low-, moderate- and high-affinity binders, respectively. So the default threshold of ANN was 0.4. CTL epitopes within HTNV-NP were predicted by ANN, together with two online algorithms BIMAS and SYFPEITHI.③The predicted HLA-A^＊02 restricted CTL epitopes were identified by ELISPOT subsequently. RESULTS： ①The prediction sensitivity of ANN, BIMAS and SYFPEITHI was 0.89, 0.78 and 0.97, respectively. The specificity was 0.90, 0.87 and 0.20, respectively. Totally 23 HTNV-NP-specific CTL epitopes were predicted by three algorithms. ②Four 9-met peptides were predicted as potential CTL epitopes within three 15-mer peptides which could stimulate PBMC from HFRS patient to secrete IFN-γ. ELISPOT assay demonstrated two of the four 9-met peptides could elicit T cell responses in PBMC from one HLA-A^＊02 positive HFRS patient.CONC