A Protein-DNA Binding Site Prediction Method Based on Multi-View Feature Fusion of Adjacent Residue

Abstract

The interaction between proteins and DNA occurs widely during the replication and transcription of DNA and other life activities. Therefore, the identification of protein- and DNA-binding sites is important for the study of protein function and drug design. Accurate prediction of binding sites has become a challenging and significant task. Although numerous studies have been conducted, prediction is challenging. In this study, a new protein-DNA binding site prediction method was proposed. This method is based on neighboring residue correlations. It uses an improved feature representation method that weighted combines several protein characteristics after a series of processing of the features and chooses a support vector machine as the prediction engine. Experiments on benchmark datasets and independent test datasets show that the proposed method has better predictability than other protein-DNA binding site predictors. This method is complementary to the existing protein-DNA binding site predictors and will be useful in the field of biotechnology.

Keywords

• Protein-DNA binding site prediction
• neighboring residue correlations
• feature processing
• multi-view features combining
• support vector machine