Reconstruction of the Protein-Protein Interaction Network for Protein Complexes Identification by Walking on the Protein Pair Fingerprints Similarity Network

Frontiers in Genetics

Zengyou He¹  Chi Lin¹  Bo Xu¹  Xiaoxia Liu²  Jie Dong²  Yu Liu³ 

[1] Key Laboratory for Ubiquitous Network and Service Software of Liaoning, Dalian, China;School of Computer Science and Technology, Dalian University of Technology, Dalian, China;School of Software, Dalian University of Technology, Dalian, China;

关键词: protein complex;    PPI network;    network reconstruction;    PPI prediction;    bioinformatics;   

DOI  :  10.3389/fgene.2018.00272

来源: DOAJ

【 摘 要 】

Identifying protein complexes from protein-protein interaction networks (PPINs) is important to understand the science of cellular organization and function. However, PPINs produced by high-throughput studies have high false discovery rate and only represent snapshot interaction information. Reconstructing higher quality PPINs is essential for protein complex identification. Here we present a Multi-Level PPINs reconstruction (MLPR) method for protein complexes detection. From existing PPINs, we generated full combinations of every two proteins. These protein pairs are represented as a vector which includes six different sources. Then the protein pairs with same vector are mapped to the same fingerprint ID. A fingerprint similarity network is constructed next, in which a vertex represents a protein pair fingerprint ID and each vertex is connected to its top 10 similar fingerprints by edges. After random walking on the fingerprints similarity network, each vertex got a score at the steady state. According to the score of protein pairs, we considered the top ranked ones as reliable PPI and the score as the weight of edge between two distinct proteins. Finally, we expanded clusters starting from seeded vertexes based on the new weighted reliable PPINs. Applying our method on the yeast PPINs, our algorithm achieved higher F-value in protein complexes detection than the-state-of-the-art methods. The interactions in our reconstructed PPI network have more significant biological relevance than the exiting PPI datasets, assessed by gene ontology. In addition, the performance of existing popular protein complexes detection methods are significantly improved on our reconstructed network.

【 授权许可】

Unknown