基于结构域信息的蛋白质复合物识别与疾病基因预测

发布时间：2018-03-25 03:06

本文选题：结构域相互作用网络　切入点：头脑风暴策略　出处：《华中师范大学》2016年硕士论文

【摘要】：蛋白质及其相互作用在各种生命活动中起着至关重要的作用。复杂的蛋白质相互作用网络中包含着许多具有研究价值的信息,亟待有效的方法对其进行深入分析与挖掘。结构域作为蛋白质中重要的空间结构层次,提供了不可或缺的重要生物信息。本文提出了一种基于智能优化思想的蛋白质复合物识别算法,并且结合蛋白质结构域相互作用网络对疾病基因进行了预测。针对蛋白质相互作用网络(PPI)拓扑结构的局限性与偏好性,本文结合结构域和拓扑结构信息对PPI网络进行了重新构建。通过与基于拓扑结构的PPI网络和结合GO注释构建的PPI网络进行比较以验证添加结构域信息后的混合结构的PPI网络有效性。基于群智能优化中的头脑风暴策略,提出了一种蛋白质复合物识别算法——IPC-BSS(Identifying Protein Complexes based on Brain Storming Strategy)。IPC-BSS算法模仿人类头脑风暴的讨论过程,设计了使蛋白质节点可以在不同蛋白质复合物之间迁移,以及相互之间联系紧密的蛋白质复合物可以进行融合的两种更新策略,以得到更优的蛋白质复合物。该算法有效克服了在早期被错误划分的蛋白质节点难以纠正的缺陷。实验结果表明,与其它经典蛋白质复合物识别算法相比,IPC-BSS算法有较好的F-measure值,能有效识别具有生物学意义的蛋白质复合物,并能完全识别一些规模较大的蛋白质复合物。蛋白质结构域之间会存在一定地联系,形成结构域相互作用网络,它从空间结构层次上解释了蛋白质之间的相互作用,本文将结构域相互网络作为蛋白质相互作用网络与疾病相似性网络的桥梁,融合多层异构网络的信息,提出了非平衡三随机游走算法——UThrRW(Unbalanced Three Random Walk)。UThrRW算法不仅挖掘了网络内的信息,同时也利用了网络之间隐藏的生物信息,结构域相互作用网络会增强蛋白质相互作用与疾病相似性网络间的联系,从而提高算法的预测效果。实验结果表明,UThrRW算法预测疾病基因的效果优于经典的RWR算法、BiRW_b1算法和UBiRW算法。
[Abstract]:Proteins and their interactions play a vital role in a variety of life activities. Complex protein interaction networks contain a lot of valuable information. It needs to be analyzed and excavated by effective methods. Domain is an important spatial structure level in protein. This paper presents a protein complex recognition algorithm based on intelligent optimization. The disease gene was predicted by protein domain interaction network. The limitation and preference of PPI topological structure of protein interaction network were analyzed. This paper reconstructs PPI network with domain and topology information, and compares it with PPI network based on topology structure and PPI network based on go annotation to verify the mixed junction after adding domain information. Based on brainstorming strategy in swarm intelligence optimization, In this paper, a protein complex recognition algorithm, IPC-BSS identification Protein Complexes based on Brain Storming Strategy).IPC-BSS algorithm, is proposed to mimic the human brainstorming process, and the protein nodes are designed to migrate between different protein complexes. And two strategies for fusion of closely connected protein complexes, In order to obtain a better protein complex, the algorithm overcomes the defects of the protein nodes that were misdivided in the early stage. The experimental results show that the IPC-BSS algorithm has a better F-measure value than other classical protein complex recognition algorithms. It can effectively recognize protein complexes with biological significance, and can fully recognize some large scale protein complexes. There will be some relationship between protein domains and form a domain interaction network. It explains the interaction between proteins from the spatial structure level. In this paper, the domain interaction network is used as a bridge between the protein interaction network and the disease similarity network, and the information of the multilayer heterogeneous network is fused. A non-equilibrium three-random walk algorithm, UThrRWbalanced Three Random Walk).UThrRW algorithm, is proposed, which not only exploits the information in the network, but also uses the hidden biological information between the networks. Domain interaction networks enhance the relationship between protein interactions and disease similarity networks. The experimental results show that UThrRW algorithm is better than the classical RWR algorithms in predicting disease genes, such as BiRW _ Stub1 algorithm and UBiRW algorithm.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：Q51;Q811.4

【相似文献】