基于空间映射的蛋白质相互作用网络链接预测算法

Link Prediction Algorithm in Protein-Protein Interaction Network Based on Spatial Mapping

蛋白质间的相互作用预测问题本质上是复杂网络的链接预测问题。到目前为止,已经有很多方法用于链接预测,这些方法要么只考虑拓扑信息,要么只考虑蛋白质相互作用网络内部的交互信息,但是仅考虑一种信息来预测蛋白质的交互信息是远远不够的。因此提出了一种新方法:将蛋白质相互作用网络看作是一个有权图,根据网络中两节点的拓扑结构和属性信息,分别计算它们的拓扑相似度和属性相似度来预测它们之间是否存在链接关系。在两种相似度平衡方面,考虑基于空间映射的方法,将它们独立地映射到另一空间,并且使它们分别映射的空间尽量相近,从而使得拓扑信息、属性信息有机融合。实验结果表明,提出的算法具有较好的准确率和良好的生物统计特性。

Protein-protein interaction(PPI)prediction is essentially the link prediction problem in the complex network.So far,many of the proposed link prediction methods either only consider topological information,or only consider the PPI interaction information within the network,but it is not enough.Therefore,this paper proposed a new method where the PPI network is represented as a weighted graph.In the graph,according to the two nodes' topology information and attribute information,the topology similarity and attribute similarity can be calculated so as to predict whether there are links between the two nodes.In order to balance the two similarities,we considered the method based on spatial mapping,that is,the similarities are independently mapped to another space,and the spaces are made as close as possible,so as to fuse the topology information and attribute information fusion.The results show that the proposed algorithm has better accuracy and good biometric characteristic.