一种新的基于多源信息的复合物识别算法

A New Method for Protein Complex Detection Based on Multi-Source Information

摘要: 蛋白质复合物是生物网络中一个重要的生物功能模块，在生物活动中起着重要的作用。通常在PPI中仅利用拓扑结构作为复合物识别的条件，约束了复合物识别的综合性能。针对此问题，本文提出一种新的基于多源信息的复合物识别算法。首先，基于GO的语义相似性构建有权平均蛋白质网络模型，然后设计以稠密子图的密度、直径和余玄相似度为聚类条件的核–附属的聚类算法；其次，根据网络中节点度值降序依次选择种子节点，并且在有权网络模型中对种子节点进行扩展以识别蛋白质复合物。最后，本文将其与CFinder，MCode和MCL进行比较和分析，实验结果表明，本文提出的算法在准确率，F度量和功能富集分析的方面均能有效地提高复合物识别的性能。

Abstract: Protein complexes are important biological function modules in protein interaction (PPI) network and play an important role in biological activities. Usually the topological structure is used as the condition for clustering in the PPI network, which constrains the overall performance of complex identification. To improve this problem, a new complex identification algorithm is proposed based on fusion of multi-source information. Firstly, based on the semantic similarity of GO, we construct a weighted PPI model and design a core-attachment clustering algorithm based on the density, diameter and cosine similarity. Secondly, the seed nodes are sequentially selected according to the descending order of the node degrees in the network and expanded in weighted graph model to identify protein complex. Lastly, the proposed method is compared with three classical algorithms. The experimental results indicate that the algorithm can effectively enhance the performance in terms of precision, F-measure and functional enrichment analysis.