Recognition of microRNA-binding sites in proteins from sequences using Laplacian Support Vector Machines with a hybrid feature

Abstract

The recognition of microRNA (miRNA)-binding residues in proteins would further enhance our understanding of how miRNAs silence their target genes and some relevant biological processes. Due to the insufficient labeled examples, traditional methods such as SVMs could not work well on such problems. Thus, we propose a semi-supervised learning method, i.e., Laplacian Support Vector Machine (LapSVM) for recognizing miRNA-binding residues in proteins from sequences by making use of both labeled and unlabeled data in this article. A hybrid feature is put forward for coding instances which incorporates evolutionary information of the amino acid sequence and mutual interaction propensities in protein-miRNA complex structures. The results indicate that the LapSVM model receives good performance with a F1 score of 22.06±0.28% and an AUC (area under the ROC curve) value of 0.760±0.043. A web server called MBindR is built and freely available at http:// cbi.njupt.edu.cn/MBindR/MBindR.htm for academic usage.