FRTpred

FRTpred predicts the logarithmic protein folding rate constant ln(k_f) and the folding type from amino acid sequences to characterize protein folding dynamics.

Key Features:

• Simultaneous Prediction: Predicts ln(k_f) and folding type concurrently from amino acid sequence data.
• Baseline Model Construction: Constructs 30 baseline models including regression models for ln(k_f) and classification models for folding type.
• Feature Extraction Diversity: Integrates 10 representative feature extraction methods across baseline models.
• Machine-Learning Algorithms: Employs three machine-learning algorithms across the baseline models.
• Ensemble Integration: Combines baseline model predictions using a random forest algorithm to produce the final predictive model.
• Validation and Performance: Cross-validation reports mean absolute deviations of 1.491 (non-two-state), 2.016 (two-state), and 1.954 (combined) for ln(k_f) and a folding type prediction accuracy of 0.843.
• Comparative Superiority: Independent testing indicates higher precision than existing methods for both ln(k_f) and folding type predictions.

Scientific Applications:

• Foldomics characterization: Accelerates characterization of foldomics datasets by providing sequence-based predictions of folding rates and types.
• Protein engineering and design: Informs protein engineering and design by supplying predicted folding kinetics and mechanisms from sequence.
• Folding-related disease research: Supports studies of folding-related diseases by predicting folding behavior associated with misfolding.
• Drug discovery and synthetic biology: Facilitates drug discovery, synthetic biology, and development of novel therapeutics through sequence-derived folding predictions.

Methodology:

Constructs 30 baseline models using 10 feature extraction methods and three machine-learning algorithms (regression for ln(k_f) and classification for folding type), then combines baseline outputs with a random forest; performance assessed by cross-validation and independent testing.