Shape string: A new feature for prediction of DNA-binding residues

Protein–DNA interactions are involved in many biological processes essential for gene expression and regulation. To understand the molecular mechanisms of protein–DNA recognition, it is crucial to analyze and identify DNA-binding residues of protein–DNA complexes. Here, we proposed a novel descriptor shape string and another two related features shape string PSSM and shape string pair composition to characterize DNA-binding residues. We employed the new features and the position-specific scoring matrix (PSSM) for modeling and prediction. The results of a benchmark dataset showed that our approach significantly improved the accuracy of the predictor. The overall accuracy of our approach reached 85.86% with 85.02% sensitivity and 86.02% specificity. The results also demonstrated that shape string is a powerful descriptor for the prediction of DNA-binding residues. The additional two related features enhanced the predictive value. ► Three new features are proposed to characterize DNA-binding residues. ► We modified the evaluation criterion of LibSVM using the G-mean metric. ► Our approach significantly improved the accuracy of the predictor.