基于模拟表位映射的B细胞表位预测方法研究

发布时间：2018-05-07 14:10

本文选题：B细胞表位 + 多肽链　；参考：《东北师范大学》2012年硕士论文

【摘要】：在免疫学领域，B细胞表位预测作为一个核心的问题在近些年被深入的讨论。B细胞表位是在抗原表面的一个可以被B细胞表面受体特异性识别并引发机体免疫的表面区域。它在疫苗研制、分子识别领域中都有很高的研究价值。然而由于鉴定表位的主要手段X射线晶体衍射和核磁共振方法周期很长、成本很高，目前只有一少部分的表位被鉴定出来。因此通过计算的方法预测表位被广泛的应用于表位鉴定中。该研究的重点是如何利用模拟表位信息准确的预测表位。在生存压力下，位于抗原表面的B细胞表位为了避免被抗体识别、特异性结合，构成表位的氨基酸进化得更快。因而与其它的蛋白质和蛋白质相互作用区域相比，表位具有较低的保守型。另外一些研究表明一个抗原表面上可能存在多个潜在的表位，但是哪个表位起作用决定于发生作用的抗体。为了解决这一复杂问题，，一种基于实验获取的多肽链的表位预测方法被提出。这些实验筛选获得的多肽链由于并认为与真实表位具有功能等价性，也称作模拟表位。因此，基于模拟表位的预测方法的主要是思路就是将模拟表位映射到抗原表面相似的区域，从而找到真实表位。模拟表位映射的方法有很多，本文尝试使用并改进了现有的图搜索模型。该算法采用了可变的距离阈值和紧凑因子来定义由抗原表面氨基酸构成的顶点间的连接，从而更好的描述抗原表面氨基酸相互作用的灵活性；优化了原有模型。此外本文还首次引入划分思想，通过表面补丁将抗原表面氨基酸划分成为一些连续的集合，从而降低了搜索的复杂度，提高了搜索的效率。在包含17个数据的测试集上，该算法与同类算法相比表现出了更高的敏感性和准确性。此外，该算法还具有最高的时间效率。在此基础之上，本文还提出了软件融合的思想来进一步提高预测的性能。通过分析判断多个软件的预测结果，集成的软件可以判断出独立的软件预测的可靠性。实验结果显示集成的软件与每个软件相比具有更高的特异性和准确性。此外，通过分析发现，预测结果的一致性与预测的可靠性呈现出明显的正相关。为了方便使用和进一步验证该方法，该集成软件被部署为一个网络服务的软件，链接为http://informatics.nenu.edu.cn。
[Abstract]:The prediction of B cell epitopes as a core issue in immunology has been discussed in depth in recent years. B cell epitopes are a surface region on the surface of antigen that can be specifically recognized by B cell surface receptors and trigger immunity. It has high research value in the field of vaccine development and molecular recognition. However, due to the long period and high cost of X-ray crystal diffraction and nuclear magnetic resonance methods, only a few epitopes have been identified. Therefore, the predicted epitopes are widely used in epitope identification. This study focuses on how to accurately predict epitopes using analog epitope information. Under survival pressure, B cell epitopes located on the surface of antigens evolve faster in order to avoid being recognized and specifically bound by antibodies. Therefore, the epitope has a lower conserved type than other protein and protein interaction regions. Other studies have shown that multiple potential epitopes may exist on the surface of an antigen, but which epitope acts depends on the antibody that acts. In order to solve this complex problem, an epitope prediction method based on experimental polypeptide chain was proposed. The polypeptide chains obtained by these experiments are also known as analog epitopes due to their functional equivalence with real epitopes. Therefore, the main idea of the analog epitope prediction method is to map the analog epitopes to the similar regions on the surface of the antigen, so as to find the real epitopes. There are many methods to simulate epitope mapping. This paper tries to use and improve the existing graph search model. In this algorithm, variable distance threshold and compact factor are used to define the connection between the peaks composed of amino acids on the surface of antigen, so that the flexibility of describing the interaction of amino acids on the surface of antigen is better, and the original model is optimized. In addition, the partition idea is introduced for the first time, and the amino acid on the surface of antigen is divided into some successive sets by surface patch, which reduces the complexity of search and improves the efficiency of search. On the test set containing 17 data, the algorithm is more sensitive and accurate than the similar algorithm. In addition, the algorithm also has the highest time efficiency. On this basis, this paper also proposes the idea of software fusion to further improve the performance of prediction. By analyzing and judging the prediction results of multiple software, the integrated software can judge the reliability of independent software prediction. Experimental results show that the integrated software is more specific and accurate than each software. In addition, it is found that the consistency of prediction results is positively correlated with the reliability of prediction. To facilitate the use and further validation of the method, the integrated software is deployed as a network service software linked to http: / / informatics.nenu.edu.cn.
【学位授予单位】：东北师范大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：R392

【相似文献】