NetTurnP

NetTurnP predicts β-turn residues and β-turn types in protein sequences using neural networks, evolutionary information, and predicted protein sequence features.

Key Features:

• Two-Class Prediction: Performs per-residue binary classification of β-turn versus non-β-turn positions.
• Type-Specific Predictions: Predicts β-turn types I, I', II, II', VIII, VIa1, VIa2, VIba, and IV using labels derived from PROMOTIF classifications.
• Input Features: Incorporates evolutionary information and predicted protein sequence features as model inputs.
• Modeling Approach: Uses neural network methodologies for prediction.
• Performance Metrics: Evaluated on the BT426 dataset of non-homologous sequences, achieving MCC 0.50 for two-class prediction, Qtotal 82.1%, sensitivity 75.6%, PPV 68.8%, and AUC 0.864.
• Comparative Advantage: Shows higher performance than eleven other β-turn prediction methods (their MCCs ranged 0.17–0.47) and attains type-specific MCCs of 0.36 for type I and 0.31 for type II.

Scientific Applications:

• Structural Biology: Enables identification of β-turn locations and types to inform protein structural characterization.
• Protein Folding Studies: Supports analyses of folding mechanisms by mapping β-turn occurrences in sequences.
• Stability Assessment: Assists evaluation of protein stability through detection of turn-related structural elements.
• Molecular Interaction Analysis: Aids interpretation of regions involved in molecular recognition and binding.
• Rational Design: Informs drug design and enzyme engineering by providing residue-level turn annotations.
• Disease Mutation Interpretation: Helps assess the potential structural impact of disease-related sequence variants on β-turns.

Methodology:

Uses neural network methodologies with evolutionary information and predicted protein sequence features as inputs, and assigns β-turn type labels based on PROMOTIF classifications.