NADbinder

NADbinder predicts NAD-interacting residues and NAD-binding proteins from amino acid sequences to support annotation of NAD binding sites and study NAD+-mediated cellular functions.

Key Features:

• SVM-Based Prediction Methodology: Employs a Support Vector Machine (SVM) classifier to distinguish NAD-interacting residues from non-interacting residues using sequence and evolutionary information.
• Position Specific Scoring Matrix (PSSM): Incorporates PSSMs generated from query sequences to capture evolutionary information, achieving a Matthews Correlation Coefficient (MCC) of 0.75 and accuracy of 87.25%.
• Non-Similarity Based Approach: Uses sequence-based prediction rather than structural similarity to identify NAD binding proteins and their interacting residues.
• Comprehensive Dataset Utilization: Models were developed on a dataset of 195 non-redundant NAD binding protein chains extracted from the Protein Data Bank (PDB).
• Cross-Validation for Model Evaluation: Model performance was assessed using five-fold cross-validation on the dataset.

Scientific Applications:

• Annotation of NAD binding proteins: Predicts NAD interacting residues (NIRs) to aid annotation of NAD binding proteins (NADBP) and their interaction sites.
• Investigation of NAD+-mediated functions: Provides residue-level predictions to support studies of metabolic and regulatory activities mediated by NAD+.

Methodology:

Analyzes amino acid sequences using PSSM-derived evolutionary features, integrates residue preference information (preferred: Gly, Tyr, Thr, His; non-preferred: Ala, Glu, Leu, Lys) into an SVM-based classifier trained on 195 non-redundant NAD-binding chains from the PDB and evaluated by five-fold cross-validation.