Prediction of protein stability changes for single-site mutations using support vector machines

Area of Science:

• Biochemistry
• Computational Biology
• Protein Engineering

Background:

• Accurate prediction of protein stability changes from single amino acid mutations is vital for protein structure understanding and protein design.
• Existing methods often require tertiary structure information, limiting their applicability.

Purpose of the Study:

• To develop and evaluate a machine learning approach for predicting protein stability changes upon single amino acid mutations.
• To assess the performance using both sequence and structural information, and to determine the feasibility of using sequence data alone.

Main Methods:

• Utilized support vector machines (SVMs) to predict stability changes.
• Leveraged both protein sequence and structural information.
• Employed cross-validation on a large dataset of single amino acid mutations.

Main Results:

• Achieved 84% accuracy in predicting the sign of stability changes, a significant improvement over prior work.
• Demonstrated that prediction accuracy using only primary sequence information is comparable to using tertiary structure information.
• The developed method is applicable even when tertiary structure is unknown.

Conclusions:

• The developed method accurately predicts protein stability changes using primary sequence information, overcoming limitations of structure-dependent methods.
• This approach enhances the ability to understand protein structure-function relationships and aids in protein design.
• The MUpro web server and associated resources are publicly available for broader research application.