面向大规模蛋白质网络功能模块检测的群智能算法研究

A Protein-Protein Interaction (PPI) network is a bimolecular relationship network which plays an important role in biological activities. Studies of functional modules in a PPI network contribute greatly to the understanding of biological mechanisms. Thus, the functional modules detection for a PPI network is an important research topic in bioinformatics in recent years, where the research which employs swarm intelligence to detect the functional modules of a PPI network has become a new hot topic in the proteomics research. However, most of the research efforts still remain at the early stage; there is almost not intensive research on the principle. In particular, the existing works ignored or weakened the deficiencies of the swarm intelligence algorithm on computing performance while analysing large PPI networks. Therefore, this project will focus on the swarm intelligent searching mechanism and its parameter optimization, which are two key scientific issues, to perform the following researches: 1) Based on the idea of problem reduction，the project merges granular computing with the swarm intelligent searching to establish a new framework that employs a multiple-grain representation model for a PPI network and a swarm intelligent optimization to effectively analyze the large-scale PPI networks, the goal of the research is to enhance the scalability of swarm intelligent detection algorithms. 2) In light of ant colony and bee colony clustering mechanisms, the project develops some novel approaches to detect functional modules in large-scale PPI networks, and looks for some new and fast algorithms with nearly linear complexity. 3) By means of the solution evolution in a population and the information exchange and co-evolution between different populations, the project studies on the parallel evolution mechanism of swarm intelligence, in which the aim is to develop some new detection methods with high performance. 4）Using association rule and Bayesian network technologies, the project establishes two parameter models for the swarm intelligent evolution in detecting functional modules of a PPI network, which will reveal the inherent relationships between different parameters and influence each other and provide a theoretical basis for the simultaneous optimization of parameters. This proposal focuses on some new attempts and explorations about swarm intelligent methods for functional module detection in large-scale PPI networks. The research will likely facilitate the development of swarm intelligence theory and its application in bioinformatics, thus has a very important theoretical significance and application value.

蛋白质网络功能模块检测是近年来生物信息学中的一项前沿课题，基于群智能机理的检测方法是新兴起的一个研究热点。然而，目前多数工作缺乏原理上深入细致的研究，尤其是忽略或淡化了群智能算法在求解大规模问题时计算性能上的不足。为此，本项目围绕群智能搜索机制及其参数优化两个关键问题进行如下研究：基于问题归约思想，将粒度计算与群智能搜索相结合，研究多粒度描述模型及其群智能优化的算法框架，旨在拓展群智能检测算法的适用范围；研究新的基于蚁群和蜂群聚类机理的检测方法，探索近似线性时间复杂度的快速算法；基于种群内的解进化和不同种群间的信息交流与协同演化机理，研究群智能并行进化机制，创建高性能的检测手段；利用关联规则和贝叶斯网技术，建立群智能进化中的参数关联模型，揭示参数间内在关系和相互影响的规律，为参数的同步优化提供理论依据。项目研究有望推动群智能理论的发展及其在生物信息学中的应用，具有重要的理论意义和应用价值。

蛋白质网络功能模块检测是近年来生物信息学中的一项前沿课题，基于群智能机理的检测方法是新兴起的一个研究热点。针对群智能算法在求解大规模问题时计算性能上的不足，本项目主要开展了如下几方面的研究：1)对从PPI网络中进行功能模块检测的计算方法进行了研究综述;2)将粒度计算与群智能搜索相结合，研究多粒度描述模型及其群智能优化的算法框架，拓展了群智能检测算法的适用范围；3)从简化算法机制的角度，研究一些新的群智能优化机理的检测方法，探索了低时间复杂度的快速算法；4) 研究群智能并行进化机制，创建了高性能的检测手段；5）为建立群智能进化中的参数关联模型，研究贝叶斯网模型学习的新方法；6）拓展了群智能学习方法在脑效应连接网络学习中的新应用。总之,项目综述了该领域的研究,并利用问题归约、低复杂度的算法设计、并行计算等思想，从方法、机制和模型多层次对群智能检测算法及相关技术进行了研究，探索了一些新方法和新途径。取得的主要研究成果包括,在国际知名刊物《IEEE Transactions on Knowledge and Data Engineering (TKDE)》（CCF A类）、《IEEE/ACM Transactions on Computational Biology and Bioinformatics》、《Information Sciences》、《BMC Bioinformatics》、《Computational Intelligence》、《International Journal of Data Mining and Bioinformatics》、《International Journal of Approximate Reasoning》、《Soft Computing》、《PLoS ONE》等上发表论文30余篇，其中，SCI期刊论文11篇，包括 CCF推荐的优秀国际刊物 8篇（1篇TKDE为顶级刊物），EI期刊论文4篇(含录用)。这些研究推动了群智能理论的发展及其在蛋白质网络、脑网络等生物信息学中的应用，具有重要的理论意义和应用价值。