Prediction of DNA-binding residues from sequence
Bioinformatics
| Hi-index | 0.00 |
Recently several machine learning approaches have been attempted to predict RNA-binding residues in amino acid sequences. None of these consider interacting partners (i.e., RNA) for a given protein when predicting RNA-binding amino acids, so they always predict the same RNA-binding residues for a given protein even if the protein may bind to different RNA molecules. In this study, we present a support vector machine (SVM) classifier that takes an RNA sequence as well as a protein sequence as input and predicts potential RNA-binding residues in the protein. The interaction propensity between an amino acid and nucleotide obtained from the extensive analysis of the representative protein-RNA complexes in the Protein Data Bank (PDB) was encoded in the feature vector of the SVM classifier. Four biochemical properties of an amino acid (the side chain pKa value, hydrophobicity index, molecular mass, and accessible surface area) were also encoded in the feature vector. On a dataset of 145 protein sequences and 78 RNA sequences, the SVM classifier achieved a sensitivity of 72.30% and specificity of 78.03%.