An accurate feature-based method for identifying DNA-binding residues on protein surfaces.

Area of Science:

• Structural biology
• Bioinformatics
• Molecular biology

Background:

• Proteins interacting with DNA are crucial for gene expression and regulation.
• Accurate identification of DNA-binding residues on protein surfaces is essential for understanding these interactions.

Purpose of the Study:

• To develop a novel computational method for predicting DNA-binding residues on protein surfaces.
• To improve the accuracy of DNA-binding residue prediction by incorporating new features.

Main Methods:

• Utilized a dataset of 119 protein-DNA complexes from the Protein Data Bank (PDB).
• Developed a prediction model using support vector machine (SVM) incorporating B-factor and packing density features alongside conventional ones.
• Evaluated the predictor using 5-fold cross-validation and analyzed features on independent holo-apo protein structure pairs.

Main Results:

• The developed SVM model accurately predicts DNA-binding residues on protein surfaces, even from DNA-free structures.
• The novel features (B-factor and packing density) significantly enhance prediction accuracy.
• The proposed method shows a significant improvement over existing approaches like DISPLAR.

Conclusions:

• The novel computational approach effectively predicts DNA-binding residues using B-factor and packing density.
• This method offers a valuable tool for studying protein-DNA interactions and guiding experimental work like site-directed mutagenesis and docking.