Extracting Sequence Features to Predict DNA-Binding Proteins Using Support Vector Machine

Abstract

DNA-binding proteins plays important role in a variety of vital biology processes. In this study, we apply a machine learning method for classify DNA-binding proteins from non-binding proteins based on sequence information. Using an evolutionary feature and residue composition feature extracted from primary structure, we have trained a support vector machine(SVM) to distinguish DNA-binding proteins from other proteins that do not binding DNA. The prediction performances are evaluated by independent test dataset which contains 361 DNA-binding proteins and 361 non-binding proteins. Our proposed method outperforms the other existing methods in the test dataset. The results achieved by our proposed method for accuracy, 84.16%; sensitivity, 85.47%; specificity, 82.89% and Matthews correlation coefficient(MCC), 0.5828 demonstrate its good performance.