PredRBR

PredRBR predicts RNA-binding residues in proteins to facilitate understanding of protein–RNA interactions involved in post-transcriptional RNA processing and transcriptional regulation.

Key Features:

• Gradient Tree Boosting Algorithm: Employs gradient tree boosting for residue-level prediction.
• Optimal Feature Set Selection: Integrates an extensive set of sequence and structural features, including two main categories of structural neighborhood properties.
• High Predictive Performance: Achieves accuracy 0.84, sensitivity 0.85, Matthews Correlation Coefficient (MCC) 0.55, and Area Under the Curve (AUC) 0.92 in cross-validation on the RBP170 dataset.
• Superior to Existing Methods: Outperforms Support Vector Machine, Random Forest, and Adaboost for RNA-binding residue prediction.
• Validated on Independent Test Sets: Demonstrates improved performance on independent test set RBP101 compared with state-of-the-art methods.

Scientific Applications:

• Functional Genomics: Elucidating gene regulation mechanisms at the post-transcriptional level.
• Structural Biology: Assisting structural characterization of protein–RNA complexes.
• Drug Discovery and Development: Identifying potential interaction sites for therapeutic targeting.

Methodology:

Uses gradient tree boosting trained on features derived from sequence and structural characteristics, including two main categories of structural neighborhood properties, with performance evaluated by cross-validation on RBP170 and testing on RBP101.