A Novel Algorithm for Prioritizing Disease Candidate Genes from the Weighted PPI Network

Computational methods accurately prioritizing latent disorder genes require for all kinds of biological information. But for the defection of a single type of biological data has a negative impact on the identification of genes causing diseases. To address the limitation, computing approaches often integrate different type of biological data. In the study, a novel algorithm PDGPC (Predicting Disease Genes with Protein Complexes) is proposed. It utilizes protein subcellular localizations to improve the reliability of the protein-protein interactions and constructs the weighted networks. And then, PDGPC builds the disease-specific networks by utilizing the protein complexes which are detected from the weighted networks through the non-negative matrix factorization. Finally, PDGPC scores all proteins in the disease-specific networks in terms of WDC. The literature retrieving method tests the correlations of top genes with more higher scores with diseases. Results show PDGPC discover some novel candidate disease genes which are valuable references for the biomedical scientists.