基于蛋白质磷酸化相关位点—修饰网络的翻译后修饰位点预测研究

发布时间：2018-01-05 07:06

本文关键词：基于蛋白质磷酸化相关位点—修饰网络的翻译后修饰位点预测研究　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：蛋白质翻译后修饰(post-translational modification,PTM)是一种十分重要的生命活动调控方式,可以改变蛋白质的结构并完善蛋白质的功能。因此深入研究蛋白质翻译后修饰的原理、种类、作用机制对于理解人类疾病的发病机制具有重要意义。近年来,随着实验技术的不断发展,积累了大量的蛋白质翻译后修饰位点数据,极大地推动了蛋白质翻译后修饰的研究进展。然而实验方法往往耗时耗力且成本较高,因此有必要发展高效、精确的计算方法预测翻译后修饰位点,为后续实验工作提供有用的参考信息。现有的计算方法大部分使用蛋白质氨基酸序列信息进行预测,忽视了蛋白质翻译后修饰间的功能联系信息。有研究表明in situ PTM指的是相同蛋白质同一个位点上发生多种翻译后修饰类型,可以反映出翻译后修饰间的功能联系。因此由多位点-多修饰相互作用关系的insituPTM启发本文从磷酸化相关位点-修饰网络的角度充分考虑网络拓扑结构信息,进而应用于翻译后修饰位点预测中来。本文主要研究内容如下:1、利用多种翻译后修饰数据库中收集的丝/苏/酪氨酸位点上的翻译后修饰位点数据,构建了磷酸化相关位点-修饰网络。在该网络的基础上,提出了基于资源配置的网络链路预测算法 SMNBI(site-modification network based inference)用于磷酸化位点预测。该算法主要利用网络中已知的链路信息对未知的位点与修饰相互作用关系进行预测。将SMNBI算法与现有的网络链路预测算法及磷酸化位点预测方法进行比较,结果表明磷酸化相关位点-修饰网络在磷酸化位点预测中发挥重要作用,能够大幅度提高预测精度。2、为解决磷酸化相关位点-修饰网络中孤立节点的预测问题,本文进一步提出了多核支持向量机算法 MK-SVM(multiple kernels support vector machine)用于全面预测丝/苏/酪氨酸位点上的多种翻译后修饰类型。首先采用高斯核函数和氨基酸置换矩阵BLOSUM62分别设计了高斯互作谱相似性核和蛋白质局部序列相似性核,然后通过线性加权组合核函数的方式输入到SVM进行训练和预测。MK-SVM算法不仅有效解决了磷酸化相关位点-修饰网络中孤立节点的预测问题,而且可对丝/苏/酪氨酸位点上的多种翻译后修饰进行预测。与多种常用的翻译后修饰预测方法的比较结果显示,该算法对于磷酸化、O-GlcNAc、硝基化、硝化等翻译后修饰类型均取得了良好的预测性能。
[Abstract]:Post-translational modification (PTM) is a very important way to regulate life activities. It can change the structure of protein and improve the function of protein. Therefore, it is very important to study the mechanism of post-translational modification of protein for understanding the pathogenesis of human disease in recent years. With the development of experimental technology, a large amount of post-translational modification site data have been accumulated, which has greatly promoted the research progress of protein post-translational modification. However, the experimental methods are often time-consuming and costly. Therefore, it is necessary to develop efficient and accurate calculation methods for prediction of post-translational modification sites. It provides useful reference information for further experiments. Most of the existing calculation methods use protein amino acid sequence information to predict. Some studies have shown that in situ PTM refers to multiple types of posttranslational modification at the same site of the same protein. It can reflect the functional relationship between posttranslational modifications. Therefore, insituPTM inspired this paper to consider the network extension from the point of view of phosphorylation related locus-modified network. Structure information. The main contents of this study are as follows: 1, using the post-translational modification site data collected from a variety of post-translational modification databases on the silk / Su / tyrosine locus. The phosphorylation related locus, modified network, was constructed on the basis of this network. A resource allocation based network link prediction algorithm, SMNBI(site-modification network based in Conference, is proposed. This algorithm mainly uses the known link information in the network to predict the interaction between the unknown sites and the modified sites. The SMNBI algorithm is combined with the existing network link prediction algorithms and phosphoric acid. The prediction methods of chemical sites were compared. The results show that phosphorylation related locus-modified network plays an important role in the prediction of phosphorylation sites and can greatly improve the prediction accuracy of .2. In order to solve the problem of prediction of isolated nodes in phosphorylation related loci-modified networks. In this paper, a new multi-kernel support vector machine algorithm, MK-SVM(multiple kernels support vector machine, is proposed. It is used to fully predict the types of post-translational modifications on silk / Sudan / tyrosine loci. First, we designed the Gao Si interaction spectrum similarity nuclei and proteins using Gao Si nuclear function and amino acid replacement matrix BLOSUM62, respectively. Local sequence similarity kernel. Then input into SVM by linear weighted combination kernel function to train and predict. MK-SVM algorithm not only effectively solves the prediction problem of isolated nodes in phosphorylated related locus-modified network. In addition, we can predict various post-translational modifications on silk / Su / tyrosine locus. Compared with other commonly used post-translational modification prediction methods, the proposed algorithm can be used for phosphorylated O-GlcNAc. After-translation modification such as nitration and nitration have achieved good predictive performance.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R3411

【参考文献】