PFP-Pred

PFP-Pred predicts protein fold patterns to support protein structure and function analysis.

Key Features:

• Ensemble Approach: Employs an ensemble of classifiers optimized using predicted secondary structure, hydrophobicity, van der Waals volume, polarity, polarizability, and dimensions of pseudo-amino acid composition.
• Optimized Evidence-Theoretic k-Nearest Neighbors (OET-KNN): Uses the OET-KNN rule as the core operation engine for each constituent classifier, leveraging evidence theory to optimize k-nearest neighbors decision-making.
• Weighted Voting Mechanism: Integrates outputs from individual classifiers through a weighted voting system to determine the final fold classification.
• Recognition of 27 Fold Patterns: Identifies true fold patterns among 27 possible configurations.
• Performance Metrics: Achieves an overall success rate of 62% on a testing dataset where most proteins have less than 25% sequence identity to training proteins, outperforming neural network (NN) and support vector machine (SVM) approaches by 6-21%.

Scientific Applications:

• Protein structure and function analysis: Provides fold predictions that aid interpretation of protein structure and inference of function.
• Proteomics and bioinformatics: Supports large-scale fold pattern annotation in proteomics and bioinformatics studies.
• Drug discovery: Facilitates target characterization and structural assessment relevant to drug discovery.
• Molecular biology research: Assists experimental design and hypothesis generation by predicting protein fold patterns.
• Biomaterials development: Informs development of novel biomaterials by providing structural predictions for protein components.

Methodology:

Combines an ensemble of classifiers that use OET-KNN as the core engine, with classifiers optimized on predicted secondary structure, hydrophobicity, van der Waals volume, polarity, polarizability, and dimensions of pseudo-amino acid composition, and integrates classifier outputs via weighted voting.