SOLart

SOLart predicts protein solubility from experimental and modeled protein structures to assess and optimize protein behavior for structural genomics, production, and aggregation studies.

Key Features:

• Solubility-dependent distance potentials: Uses solubility-dependent distance potentials to analyze residue-residue interactions that influence solubility.
• Structural and sequence features: Integrates structural features such as backbone torsion angles and solvent accessibility with sequence-based attributes for prediction.
• Random forest model: Employs a random forest trained on Escherichia coli proteins with known experimental structures and solubility values.
• Validation performance: Shows a Pearson correlation of approximately 0.7 between predicted and experimental solubility values in cross-validation and independent tests on Saccharomyces cerevisiae proteins.
• Resolution robustness: Maintains predictive performance across both high-resolution and low-resolution protein structures.
• Folding free energy features: Identifies folding free energy differences derived from statistical potentials as among the most informative features.

Scientific Applications:

• Structural Genomics: Assists selection of proteins likely to be soluble for structural determination efforts.
• Protein Production: Supports identification and optimization of soluble protein variants for biotechnological production.
• Disease Research: Aids investigation of proteins prone to aggregation relevant to diseases involving misfolded proteins.

Methodology:

Constructs solubility-dependent statistical distance potentials from structural features (including backbone torsion angles and solvent accessibility), combines these with sequence-based attributes, and uses them as input features for a random forest model trained on experimental solubility data; folding free energy differences derived from the statistical potentials are used as key predictive features.