利用聚类后PCA方法的T细胞表位预测研究

发布时间：2018-04-02 11:23

本文选题：T细胞　切入点：表位预测　出处：《东北师范大学》2013年硕士论文

【摘要】：T淋巴细胞(T Lymphocyte)简称T细胞。在免疫应答过程中，T细胞不能像抗体一样直接识别完整的天然抗原，而是需要借助T细胞表面的T细胞抗原受体(T-cell Receptor,TCR)识别抗原蛋白中一段具有特殊功能的肽段，这段肽段称为T细胞表位(Epitope)，或抗原决定簇(Antigenic Determinant)。主要组织相容性复合体(Major HistocompabilityComplex, MHC)根据其编码及其参与抗原提呈途径的不同可分为MHC I类分子以及MHC II类分子。分别参与内源性以及外源性抗原提呈途径。定位与MHC分子相结合的T细胞表位不仅有助于了解自身免疫性疾病、过敏反应、传染性疾病以及肿瘤等疾病的免疫应答机理，而且对于计算机辅助设计人工疫苗以及免疫干预治疗等都有非常重要的意义。目前T细胞表位的定位方法都需要消耗大量人力物力资源，，并且对设备的要求也很高。随着已知表位数据的大量增加，计算机作为一种有效的辅助实验手段渐渐被研究者应用于生物实验当中。通过计算机方法预测得到候选表位可以通过后续的生物实验加以验证。恰当地使用这种计算机与生物实验相结合的方式既可以保证结果的准确性，又可以节约成本、提高工作效率，从而满足现代社会的生产需求。目前比较成熟的MHC分子亲和肽预测方法主要有结合基序法(Binding Motif)、定量矩阵法(Quantitative Binding Matrix)以及机器学习法(Machine Learning)。本文在机器学习方法的基础上提出一种聚类后PCA方法用以约减参与MHC分子亲和肽预测的氨基酸理化性质，此方法针对不同MHC分子筛选不同的氨基酸理化性质，使其更具有针对性，以提高机器学习算法的性能。试验证实，本文的算法在准确率以及敏感性方面都有不同程度的提高，因此本文在T细胞表位预测理论研究以及实际应用方面都有推动作用。
[Abstract]:T-lymphocyte T Lymphocyte.in the course of immune response, T cells cannot recognize intact natural antigens as directly as antibodies. Instead, we need to use T-cell receptor (T-cell receptor) on the surface of T cells to recognize a peptide that has a special function in an antigen protein. The peptide segment is called T cell epitope Epitopeus, or antigenic determinant. Major histocompatibility complex Major Histocompactability Complexes (MHCs) can be divided into MHC class I and MHC II according to their coding and the way in which they participate in antigen presentation. Localization of T cell epitopes combined with MHC molecules is not only helpful for understanding autoimmune diseases. The immune response mechanism of allergic reactions, infectious diseases and tumours, It is very important for computer-aided design of artificial vaccine and immune intervention therapy. At present, the method of T cell epitope localization needs a lot of manpower and material resources, and the requirement of equipment is also very high. With the large increase in known epitope data, As an effective auxiliary experimental method, computer has been gradually applied in biological experiments. The candidate epitopes predicted by computer method can be verified by subsequent biological experiments. The combination of computer and biological experiments can guarantee the accuracy of the results. It can also save cost and improve work efficiency to meet the production needs of modern society. At present, mature MHC molecular affinity peptide prediction methods mainly include binding motif method, quantitative Binding matrix method and machine learning method. In this paper, a post-cluster PCA method is proposed based on machine learning method. The physicochemical properties of amino acids predicted by MHC molecular affinity peptide were reduced. In order to improve the performance of machine learning algorithm, different amino acid physicochemical properties were screened by this method for different MHC molecules. The experimental results show that the proposed algorithm can improve the accuracy and sensitivity of T cell epitopes in different degrees, so this paper can promote the theoretical research and practical application of T cell epitope prediction.
【学位授予单位】：东北师范大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：R392;TP311.13

【参考文献】