SABLE

SABLE predicts protein conformational flexibility and phosphorylation propensity from amino acid sequence to support analysis of structural dynamics and post-translational modification.

Key Features:

• Conformational flexibility prediction: Predicts X-ray structure-derived temperature factors (B-factors) and solvent accessibility standard deviations (SASDs) from sequence data.
• Epsilon-insensitive support vector regression (ε-SVR): Applies ε-SVR machine learning to map sequence-derived features to real-valued structural properties.
• Integration of structural propensities: Incorporates real-valued secondary structure and relative solvent accessibility (RSA/SS) predictions into predictive models.
• Position-specific scoring matrices (PSSMs): Uses PSSMs derived from multiple sequence alignments as sequence-based representations for prediction.
• Phosphorylation prediction enhancement: Combines predicted conformational flexibility measures with PSSMs to improve phosphorylation site prediction.
• Classification approaches for phosphorylation: Evaluates one-class and two-class support vector machine (SVM) classifiers to address unknown negative examples in training data.

Scientific Applications:

• Protein structure analysis: Provides B-factor and SASD predictions to inform studies of protein dynamics and stability.
• Functional annotation: Enhances prediction of phosphorylation sites to support annotation of regulatory and signaling functions.
• Drug discovery and design: Supplies information on conformational flexibility and modification propensity relevant for target characterization and ligand design.

Methodology:

Uses sequence-based representations including PSSMs from multiple sequence alignments and real-valued RSA/SS predictions, applies epsilon-insensitive support vector regression (ε-SVR) for real-valued property prediction, and evaluates one-class and two-class SVM classifiers for phosphorylation prediction.