支持向量机方法在T细胞表位预测中的应用

发布时间：2018-05-22 09:00

本文选题：T细胞表位 + 抗原加工提呈途径　；参考：《大连理工大学》2009年博士论文

【摘要】： 在T细胞介导的特异性免疫应答中,T细胞表面抗原受体(T-cell receptor,TCR)仅能识别抗原肽与主要组织相容性复合体(major histocompatibility complex,MHC)分子结合形成的复合物。此复合物的形成是依赖于抗原的加工提呈途径。内源性抗原(病毒、肿瘤抗原等)需要经过蛋白酶体(proteaome)的降解、与抗原提呈相关的转运蛋白(transporter associated with antigen processing,TAP)的转运、MHCⅠ类分子的结合后才能被细胞毒性T细胞(cytotoxic T lymphocyte,CTL)识别,相应的抗原肽称为CTL表位;外源性抗原(细菌产生的毒素)也需要经过溶酶体酶的降解和MHCⅡ类分子的结合后才能被辅助性T细胞(helper T cell,Th)识别,相应的抗原肽称为Th表位。一般来讲,抗原的加工提呈途径决定了T细胞对表位的选择性。为了进一步弄清抗原的加工提呈机制,提高T细胞表位预测的准确性和合理性,本文应用支持向量机方法对抗原加工提呈途径中三个重要的选择性阶段进行了理论预测研究。 1.在内源性抗原的加工提呈途径中,真核细胞中的泛素-蛋白酶体系统对抗原蛋白发挥着重要的酶切降解功能。为了进一步理解蛋白酶体的酶切机制,本文对蛋白酶体的裂解位点特异性进行了研究。文中采用支持向量分类器(support vector classifier,SVC)方法建立了蛋白酶体的裂解位点预测模型,预测准确度达到83.1%。在相同检验集下,本模型的性能表现要优于其他预测模型。通过分析预测模型中不同位置上氨基酸对裂解位点形成的权重系数,本文获取了蛋白酶体裂解位点及其两侧区域氨基酸的裂解特异性。这些裂解特异性反映了蛋白酶体与抗原蛋白相互作用信息,同时表明蛋白酶体对抗原蛋白的酶切处理不是随机的,而是有一定模式和选择性的。研究结果为进一步揭示蛋白酶体裂解抗原蛋白的机理提供了重要的信息。 2.在内源性抗原的加工提呈途径中,MHCⅠ类分子发挥着启动和调节免疫应答的重要作用。抗原肽只有结合MHCⅠ类分子后,才能被细胞毒性T细胞(CTL)识别,然而,对于一个给定的MHCⅠ类分子,只有一组特定的抗原肽才能与之结合。因此,准确判断哪些抗原肽能与指定MHCⅠ类分子结合,不仅有助于理解免疫机制,而且有助于开发高效的抗肿瘤疫苗。为了进一步了解MHCⅠ类分子与抗原肽结合的特异性,本文利用支持向量回归机(support vector regression,SVR)方法以及四种氨基酸编码方式建立了四个抗原肽与MHCⅠ类分子结合亲和力的预测模型。四个模型的性能比较显示,基于氨基酸物理化学性质建立的模型具有更好的预测能力。此外,本文通过分析抗原肽中不同位置氨基酸对结合MHCⅠ类分子形成的权重系数,获取了抗原肽与MHCⅠ类分子的结合特异性。 3.在外源性抗原的加工提呈途径中,抗原肽与MHCⅡ类分子的结合是激活辅助性T细胞特异性免疫应答的先决条件。对于给定的一种MHCⅡ类分子,准确预测与之结合的抗原肽,不仅有助于人们进一步理解免疫的基本原理,还对表位疫苗的开发、自身免疫性疾病(如类风湿关节炎、胰岛素依赖性糖尿病等)的治疗等有着重要的意义。本文应用迭代自洽(iterative self-consistent,ISC)策略与支持向量回归机(SVR)的组合方法和四种氨基酸编码方式,对17种MHCⅡ分子(包括14种人类的HLA DR分子和3种鼠类的H2 IA分子)的配体数据进行了回归分析,分别建立了预测模型。与其他预测模型的比较结果显示,本文模型具有更优的性能表现。此外,本文以HLA DRB1*0101为例,通过分析抗原肽中不同位置氨基酸对结合MHCⅡ类分子形成的权重系数,获取了抗原肽与MHCⅡ类分子的结合特异性。研究结果为进一步揭示Th细胞表位的产生机制提供了重要的信息。
[Abstract]:In the specific immune response mediated by T cells, the T cell surface antigen receptor (T-cell receptor, TCR) only recognizes the combination of the antigen peptide and the major histocompatibility complex (major histocompatibility complex, MHC) molecules. The formation of this complex is based on the process of antigen processing. Tumor antigen, etc., need to be degraded by proteasome (proteaome), the transport of transporter associated with antigen processing (TAP) associated with antigen, and the binding of MHC class I molecules can be identified by the cytotoxic T cells (cytotoxic T lymphocyte), and the corresponding antigen peptide is called the epitope; Bacteria produced toxins also need to be identified by the degradation of lysosomal enzymes and the binding of MHC class II molecules to be identified by helper T cell (Th), and the corresponding antigen peptides are called Th epitopes. Generally speaking, the process of antigen processing determines the selectivity of the epitopes of T cells. In order to further clarify the mechanism of antigen processing, In order to improve the accuracy and rationality of the prediction of T cell epitopes, this paper uses the support vector machine method to predict the three important selective stages of the original processing approach.
1. in the processing approach of endogenous antigen, the ubiquitin proteasome system in eukaryotic cells plays an important role in enzymatic degradation of antigen protein. In order to further understand the proteasome mechanism of proteasome, the specificity of proteasome cleavage site is studied in this paper. The support vector classifier (support vector) is used in this paper. Classifier, SVC) method established the prediction model of proteasome cleavage site, and the prediction accuracy reached 83.1%. in the same test set. The performance of the model was better than the other prediction models. By analyzing the weight coefficient of the amino acid on the cracking site in different positions in the prediction model, the proteasome cracking site was obtained. The cleavage specificity of the amino acids on the two sides of the region reflects the interaction information between the proteasome and the antigen protein, and indicates that the protease removal of the antigen protein is not random, but has a certain pattern and selectivity. The result of the study is to reveal the mechanism of proteasome cracking antigen protein in a further step. Important information is provided.
2. in the processing and presenting pathway of endogenous antigen, MHC class I molecules play an important role in starting and regulating the immune response. Only after the binding of MHC I molecules can the antigen peptide be identified by the cytotoxic T cells (CTL). However, only a set of specific antigen peptides can be combined with a given MHC class I molecule. Therefore, the accuracy of the antigen peptide is accurate. Judging which peptides can be combined with designated MHC class I molecules not only helps to understand the immune mechanism, but also helps to develop efficient antitumor vaccines. In order to further understand the specificity of the binding of MHC class I molecules to antigenic peptides, this paper uses the support vector regression (support vector regression, SVR) method and four amino acids coding. The prediction model of the binding affinity of four antigen peptides with MHC I was established. The performance comparison of the four models showed that the model based on the physicochemical properties of amino acids had better prediction ability. In addition, the weight coefficients of different amino acids in the antigen peptides for the formation of MHC I molecules were obtained. The binding specificity of the antigen peptide to the MHC class I molecules.
3. in the processing and presenting approach of exogenous antigen, the binding of antigen peptide to MHC class II molecules is a prerequisite for activating the specific immune response of auxiliary T cells. For a given MHC class II molecule, the accurate prediction and combination of antigenic peptides not only help people to understand the basic principles of immunization further, but also to the opening of epitope vaccines. The treatment of autoimmune diseases such as rheumatoid arthritis (rheumatoid arthritis, insulin dependent diabetes) is of great significance. In this paper, the combination of iterative self-consistent (ISC) strategy and support vector regression machine (SVR) and four amino acid coding methods are applied to the 17 MHC II molecules (including 14 kinds of human HLA DR points). The regression analysis of the ligand data of the H2 IA molecules of the 3 species of rodents and the other prediction models showed that the model had better performance. In addition, this paper took HLA DRB1*0101 as an example to analyze the weight of different positions of amino acids in the antigen peptide to the formation of MHC class II molecules. The binding specificity of antigen peptide with MHC class II molecules has been obtained. The results provide important information for further revealing the mechanism of Th cell epitopes.
【学位授予单位】：大连理工大学
【学位级别】：博士
【学位授予年份】：2009
【分类号】：R392

【引证文献】