Three Simple Properties Explain Protein Stability Change upon Mutation.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Protein Engineering

Background:

• Accurate prediction of protein stability changes upon mutation is crucial for protein engineering and understanding genetic diseases.
• Existing prediction tools range from complex energy-based models to machine learning, often lacking chemical interpretability.
• There is a need for simple, interpretable models that retain predictive power.

Purpose of the Study:

• To identify a minimal, interpretable model for predicting protein stability changes.
• To evaluate the performance of a simple multilinear regression model against complex existing methods.

Main Methods:

• Developed SimBa, a simple multilinear regression model.
• Trained SimBa on a balanced experimental dataset of amino acid substitutions.
• Input features included solvent accessibility, volume difference, and polarity difference.

Main Results:

• SimBa achieved performance comparable to more complex, widely used protein stability prediction methods.
• The model's simplicity allows for direct interpretation of chemical factors influencing stability.
• Identified a potential accuracy limit of ~1 kcal/mol for current approaches.

Conclusions:

• Simple models incorporating key chemical features can be highly effective for protein stability prediction.
• Further improvements in accuracy (e.g., to 0.5 kcal/mol) require incorporating additional factors like unfolded states and water interactions.