Bipartite Heterogeneous Network Method Based on Co-neighbor for MiRNA-Disease Association Prediction

Conclusion Inspired by the general network co-neighbors, this paper proposed the definition of the co-neighbors of the bipartite network based on the hypothesis that that functionally similar miRNAs are related to phenotypically similar diseases. Eight local structural similarity indexes which are co-neighbor, Salton, Jaccard, Sørensen, HPI, HDI, LHN1, and PA were used to measure the association probabilities between nodes. Several types calculation methods of computational miRNA-disease prediction score were introduced, namely, the bipartite network co-neighbor link prediction score using only known association information, the co-neighbor link prediction score based on miRNA similarity, the co-neighbor link prediction score based on disease similarity, the weighted co-neighbor link prediction score based on miRNA similarity and disease similarity. Using only known association information, the co-neighbor link prediction score on bipartite network cannot predict the isolated diseases and new miRNAs, but the score calculation is simple, and only the experimentally verified miRNA-disease association information can be used for inference prediction. The co-neighbor link prediction score based on miRNA similarity used the association probability of all miRNAs and specific diseases to measure the degree of association between specific miRNAs and specific diseases. Using this score can significantly improve the prediction accuracy, but it cannot be used to predict isolate...
Source: Frontiers in Genetics - Category: Genetics & Stem Cells Source Type: research