Prediction of RNA-binding sites from evolutionary information of protein sequences

Protein-RNA interactions play significant roles in a number of biological activities, such as protein synthesis, regulation of gene expression. A reliable identification of RNA-binding sites in proteins is important to understand the molecular details of protein-RNA interaction. In this work, we have developed a machine learning approach, support vector machine (SVM), to predict RNA - binding sites in proteins based on the profile of evolutionary conservation of sequence positions, which only needs protein primary sequence as input of classifier. Using evolutionary information in terms of a position specific scoring matrix (PSSM) of each residue and 6 of its closest neighboring residues, our results indicated that RNA - binding residue can be predicted at 67.6% sensitivity, 75.0% specificity and a net prediction (an average of sensitivity and specificity) of 71.3%. This method outperforms previous protein RNA - binding site prediction methods.