RNABindR

RNABindR predicts RNA-binding residues in proteins from sequence or structure to support analysis of protein–RNA interactions.

Key Features:

• Identification in solved complexes: Identifies RNA-contacting amino acids in Protein Data Bank (PDB) structures using a distance cutoff and generates labeled amino acid sequences indicating interface residues.
• Sequence-based prediction: Uses a Naive Bayes classifier trained on a non-redundant set of protein–RNA complexes from the PDB to predict RNA-binding residues from protein sequence.
• Accuracy and calibration: Achieves over 85% accuracy for interface-residue identification in leave-one-out cross-validation experiments and can be calibrated for higher specificity or higher sensitivity.
• Comparative performance: Performs comparably to neural network classifiers for sequence-based RNA-binding site prediction while providing advantages in speed, simplicity, and interpretability.
• Experimental validation: Predictions have been validated against experimental data, including studies of the human telomerase protein hTERT.

Scientific Applications:

• Experimental design: Guides mutagenesis and binding assays by predicting residues likely to contact RNA.
• Mechanistic studies: Supports analysis of the diversity, mechanisms, and regulation of RNA–protein complexes.
• Therapeutic target identification: Aids identification of residues and regions for strategies aimed at modulating RNA–protein interactions.

Methodology:

Identifies contact residues in PDB complexes using a distance cutoff; applies a Naive Bayes classifier trained on a non-redundant PDB-derived set for sequence-based prediction; performance assessed by leave-one-out cross-validation.