Plant miRNA function prediction based on functional similarity network and transductive multi-label classification algorithm

• Weight of protein–protein interaction network is calculated by using the similarity of GO terms.
• MiRNA functional similarity network is constructed by rank-based method.
• Prediction method of miRNA functions based on similarity network and transductive multi-label classification is proposed.

Plant miRNAs play critical roles in the response to abiotic and biotic stress. The advancement in the number of plant miRNA functions lags far behind that of plant miRNAs. In this paper, a method to predict the functions of plant miRNAs is proposed. The functional similarity between each pair of miRNAs is inferred based on a weighted protein–protein interaction network (WPPIN) and graph-theoretic properties. A miRNA functional similarity network (MFSN) is constructed by a simple but robust rank-based approach. Transductive multi-label classification (TRAM) is applied to the MFSN. The experimental results demonstrate that our prediction approach obtains high effectiveness in Arabidopsis thaliana. It can also be applied to other plant species when protein–protein interaction networks of various organisms are available.