蛋白质网络中的功能模块挖掘算法与关键蛋白识别方法研究

【作者】 王飞；

【导师】 雷秀娟；

【作者基本信息】 陕西师范大学 ， 计算机应用技术， 2016， 硕士

【摘要】 当前的生命科学研究正处在后基因组时代,通过信息技术手段对蛋白质进行分析研究是蛋白质组学的一个重要内容,尤其是从蛋白质相互作用网络中挖掘出有生物意义的蛋白质功能模块以及识别出生命体的关键蛋白,对揭示生命体特定活动具有重要的研究意义。蛋白质相互作用网络中的数据来自各种高通量技术手段测量得来,由于细胞环境的复杂性以及手段的局限性,得到的数据不可避免的存在着错误信息。并且已有的蛋白质相互作用数据只能显示相互作用曾在细胞存在过,至于何时存在何时消失,则无法给出。这就影响到了从中挖掘蛋白质功能模块的准确性。关键蛋白是对生命体功能的正常表达起到重要作用的蛋白质。一个关键蛋白的缺失就可能引起生命体的功能缺失。因此,关键蛋白的识别工作也引起了研究人员的密切注意。本文主要工作如下：(1)提出了将烟花算法优化马尔可夫聚类的膨胀参数用于挖掘蛋白质功能模块,并提出了将烟花算法在蛋白质相互作用网络中模拟爆炸过程直接挖掘蛋白质功能模块。第一种方法通过烟花算法动态调节马尔可夫聚类的参数达到优化聚类结果的目的,第二种方法在蛋白质网络中将烟花与周围符合条件的火花聚集成一类。实验结果表明,两种方法都可以获得较好的结果。(2)提出了将萤火虫优化算法优化马尔可夫聚类的膨胀参数用于挖掘蛋白质功能模块。通过将萤火虫优化算法用于优化马尔可夫聚类的参数,算法运行效率较之烟花算法更高。实验结果表明该方法比其他群智能优化算法以及其他功能模块挖掘算法都能获得更加准确的聚类结果。(3)提出了在蛋白质功能模块中利用它们的拓扑特性进行关键蛋白识别。在蛋白质功能模块中,针对模块子网进行拓扑特性研究,分别研究了使用度中心性和边聚集系数中心性的方法进行关键蛋白识别的优劣。实验结果表明,两种方法在大部分样本中都可以得到比其他传统方法更好的识别准确性,可以识别出更多的关键蛋白。更多还原

【Abstract】 Since the life science research is in the post-genomic era, an important content of proteomics is study and analysis the proteins by using information technology. Especially, the most important part is detecting protein functional modules and identifying essential proteins from protein-protein interaction networks, which can help to reveal the specific activities of organism.Because of the complex cell environment and the limitations of high-throughput techniques, the data in Protein-Protein Interaction (PPI) databases contains a lot of error messages. Furthermore, the protein interaction data can only show that the interaction was existed, but it cannot show when and where was it synthesized and resolved. The incomplete data has affected the accuracy of detecting protein functional modules from it. Essential protein is a kind of protein that plays an important role in organism’s function. Maybe a miss of an essential protein can cause a miss of function. So identifying essential proteins have caused researchers’attention.The main studies of this paper are as follows:Firstly, this paper uses firework algorithm (FWA, for short) to optimize the expansion parameter of Markov clustering (MCL, for short) in order to detect functional modules in protein-protein interaction networks. There are two different processes, one is simulating explosions in the network, in this way, a firework and some special sparks can form a cluster, the other is use FWA as a part of MCL to optimize a parameter of it, the they can get results together. The experiments show that both of them can get a good performance.Secondly, this paper uses firefly optimization algorithm (FA, for short) to optimize the expansion parameter of MCL in order to detect functional modules in protein-protein interaction networks. FA is more effective than other swarm intelligence optimization algorithms, and the experiments illustrate this point. The experiments also show that it can get a better performance than other common used clustering algorithms.Thirdly, this paper proposes a method to identify essential proteins in functional modules, through analyzing the topological properties of functional modules. In functional modules, this paper uses degree centrality and edge clustering coefficient centralities to analyze the topological properties. The bigger a node’s centralities, the more likely it will be an essential protein. The experiments show that this method can get more essential proteins in most samples.更多还原