NCBRPred

NCBRPred predicts DNA- and RNA-binding residues in proteins using a multilabel sequence-labeling model to identify nucleic acid binding sites for protein-nucleic acid interaction analysis.

Key Features:

• Multilabel Learning Framework: NCBRPred employs a multilabel learning approach to predict multiple types of nucleic acid binding sites within a single protein, enabling simultaneous DNA-binding and RNA-binding residue prediction.
• Sequence Labeling Model with BiGRUs: The method uses a sequence labeling model incorporating bidirectional Gated Recurrent Units (BiGRUs) to capture global interactions among amino acid residues along the protein sequence.
• Low Cross-Prediction Rate: The model minimizes misclassification between DNA-binding and RNA-binding residues, reducing cross-prediction errors.
• Benchmark Performance: Experimental validation on three widely used benchmark datasets and an independent dataset showed that NCBRPred outperforms 10 existing predictors, with higher predictive accuracy and a significantly reduced cross-prediction rate.

Scientific Applications:

• Gene Expression Research: Predicting nucleic acid binding residues facilitates analysis of protein interactions that influence gene expression.
• Protein Function Analysis: Identifying residues involved in nucleic acid interactions aids inference of protein function in cellular processes.
• Drug Design and Development: Mapping potential nucleic acid-binding sites provides targets for designing molecules that specifically modulate protein-nucleic acid interactions.

Methodology:

Uses bidirectional Gated Recurrent Units (BiGRUs) within a multilabel sequence-labeling model to capture and analyze global residue interactions.