BindN

BindN predicts DNA- and RNA-binding residues from protein primary amino acid sequences using support vector machines to identify nucleic-acid interaction sites.

Key Features:

• Input/Output: Accepts an amino acid sequence and outputs residue-level predictions of DNA or RNA binding.
• Machine Learning Approach: Uses support vector machines (SVMs) for classification of binding versus non-binding residues.
• Training Data: Models are trained on protein datasets with known DNA- or RNA-binding residues sourced from the Protein Data Bank (PDB).
• Feature Encoding: Encodes residues using side chain pK(a) value, hydrophobicity index, and molecular mass.
• Performance Metrics: Reported sensitivity and specificity are: DNA-binding residues — sensitivity 69.40%, specificity 70.47%; RNA-binding residues — sensitivity 66.28%, specificity 69.84%.
• Comparative Performance: The SVM models are reported to have higher accuracy than previous studies.

Scientific Applications:

• Gene regulation studies: Identifying nucleic-acid interaction sites to inform analyses of protein roles in gene regulation and expression.
• Protein annotation: Aiding annotation of protein functions within genomic and proteomic datasets by predicting binding residues.
• Experimental design: Prioritizing residues for experimental validation of predicted protein–nucleic acid interactions.

Methodology:

Support vector machines trained on PDB-derived protein datasets with known DNA/RNA-binding residues; residues encoded by side chain pK(a), hydrophobicity index, and molecular mass; accepts amino acid sequence input and produces residue-level binding predictions.