Protein link prediction algorithm fusing community structure and node degree local path similarity

Abstract

The invention discloses a protein link prediction algorithm fusing community structure and node degree local path similarity. The purpose of the algorithm is to predict potential links in a protein interaction network. The method mainly comprises the following steps: constructing an adjacent matrix; detecting a community structure; constructing a training set and a test set; calculating communitycompactness indexes; calculating local path similarity based on the node degree; and calculating similarity values among all the nodes which are not linked. Existing link prediction methods mostly utilize common neighbor information of nodes, and do not consider contribution of protein community structure information to link prediction. Therefore, the defects of a method based on local informationsimilarity and a protein interaction prediction method based on a community structure are considered; the invention provides a protein link prediction method fusing community structure and node degree local path similarity, and link prediction is carried out in combination with a topological structure between proteins while community information in a protein interaction network is considered.

Description

technical field [0001] The invention belongs to the field of complex networks, in particular to a protein link prediction algorithm that integrates community structure and local path similarity of node degree. Background technique [0002] In recent years, scholars at home and abroad have done a lot of work on the link prediction of protein-protein interaction network (PPI). Existing link prediction methods usually use node local information for prediction. As a classic link prediction method, methods based on local information similarity have been applied to link prediction in protein interaction networks due to their high accuracy and low complexity. Methods based on local information similarity usually perform link prediction based on the assumption that the higher the similarity between nodes, the higher the possibility of links appearing. The classic local similarity methods include Common neighbors (CN), Adamic-Adar (AA), Resource allocation (RA) and Preferential att...

AI Technical Summary

Problems solved by technology

The above methods pay more attention to mining the network topology, and lack of mining the topological information of the protein itself.

Images