基于PPI网络的关键蛋白质识别方法研究及应用

发布时间：2018-08-18 19:21

【摘要】：蛋白质是组成生物体的一切细胞、组织的重要成分,是生命的物质基础。蛋白质参与了机体的所有重要组成部分。一般来说,将蛋白质经过基因剔除式突变并将其移除后造成了生物体的功能丧失,并使生物体致病甚至无法生存,该蛋白质即为关键蛋白质。由于生物体的存活和后代的繁衍都离不开关键蛋白质,因此生命科学中一项重要的研究内容就是识别关键蛋白质。本文在蛋白质相互作用(PPI)网络拓扑结构的基础上,将融合多源生物信息来预测关键蛋白质,主要内容包括以下三个方面:(1)提出了基于改进的PageRank算法EPP(Essential Proteins Predict)识别关键蛋白质。该算法将PPI网络看成是一个不确定的、顶点带有属性的网络,然后在该网络中将重要性排名在前p%的顶点作为关键蛋白质。该方法首先需要计算顶点(即蛋白质)间的相似度,对于相似度的计算,我们综合考虑了蛋白质的可信度及语义相似度信息;其次,对于PPI网络中的每一个顶点我们还考虑了顶点的邻居信息,即计算它的邻域相似度;最后利用上述提出的可信度及语义相似度来计算顶点的重要性。本文提出的算法综合考虑了PPI网络的拓扑信息和蛋白质的生物信息,因此具有复杂度低、识别准确率高的优点。我们利用标准数据集进行测试,实验结果表明,所提出的算法能更准确地识别出更多的关键蛋白质。(2)提出了基于改进的PSO算法EPPSO(EssentialProtein PSO)识别关键蛋白质。在该算法中,我们提出了衡量top-p关键蛋白质的整体性指标,而不是评估蛋白质关键性的单个指标。EPPSO算法采用选取候选解的方法,每一个候选解含有尸个蛋白质,我们整体度量这P个蛋白质的关键性即可。对于整体关键性,我们采用这些蛋白质与其他蛋白质间的联系的紧密度来衡量,即为算法中提出的适应度函数。然后根据粒子群的算法思想,通过跟踪全局最优值及个体最优值来更新该函数。为了评估算法的性能,我们在酵母数据集等标准数据集上运行该算法,实验结果表明与其他经典算法相比,本文算法识别准确率明显优于其他方法。此外,由于本文算法只需识别出P个关键蛋白质即可,而不必根据某种指标逐个计算每个蛋白质的关键性,因此具有较低的计算量。(3)在以上工作的基础上,本文设计了一个基于WEB的在线关键蛋白质识别系统。该系统可以把预测的结果通过图形化的方式体现在系统上,方便高效。经测试该系统运行稳定,界面美观,具有良好的经济价值和社会价值。
[Abstract]:Protein is all the cells that make up the organism, the important component of the tissue, is the material foundation of life. Proteins participate in all important parts of the body. In general, the protein is a key protein because of the loss of the function of the organism and even the inexistence of the organism after the gene mutation and removal of the protein. Because the survival of organisms and the reproduction of offspring can not be separated from key proteins, one of the important research contents in life science is to recognize key proteins. In this paper, based on the topological structure of protein interaction (PPI) network, the key proteins are predicted by fusion of multi-source biological information. The main contents are as follows: (1) an improved PageRank algorithm based on EPP (Essential Proteins Predict) is proposed to recognize key proteins. The algorithm regards the PPI network as an uncertain network with attributes on the vertices, and then regards the top p% of the important vertices as the key protein in the network. The method first needs to calculate the similarity between the vertices (i.e. protein). For the similarity calculation, we consider the reliability and semantic similarity information of the protein. Secondly, For each vertex in PPI network, we also consider the neighbor information of vertex, that is, calculate the neighborhood similarity of vertex, and calculate the importance of vertex by using the credibility and semantic similarity proposed above. The proposed algorithm takes into account the topological information of PPI network and the biological information of protein, so it has the advantages of low complexity and high recognition accuracy. We use the standard data set to test, and the experimental results show that the proposed algorithm can recognize more accurate key proteins. (2) based on the improved PSO algorithm, EPPSO (EssentialProtein PSO) is proposed to recognize the key proteins. In this algorithm, we propose a holistic index to measure the key protein of top-p, rather than a single index to evaluate the key protein. EPPSO adopts the method of selecting candidate solutions, each candidate contains a protein. We can measure the key of the P protein as a whole. For the whole key, we measure the closeness of the relationship between these proteins and other proteins, that is, the fitness function proposed in the algorithm. Then according to the idea of particle swarm optimization, the function is updated by tracking the global optimal value and the individual optimal value. In order to evaluate the performance of the algorithm, we run the algorithm on standard data sets such as yeast datasets. The experimental results show that the recognition accuracy of this algorithm is better than that of other classical algorithms. In addition, because the algorithm only needs to identify P key proteins, but not to calculate each key protein one by one according to a certain index, it has a low computational complexity. (3) on the basis of the above work, An online key protein recognition system based on WEB is designed in this paper. In this system, the predicted results can be graphically reflected in the system, which is convenient and efficient. After testing, the system runs stably, the interface is beautiful, and has good economic value and social value.
【学位授予单位】：扬州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q51;O157.5

【相似文献】